{% elsif post.venue %} {{ post.venue }} {% endif %} - ({{ post.date | date: "%Y" }}) + ({{ post.date | date: "%Y" }}){% if post.award %} · {{ post.award }}{% endif %} {% endif %} + {% include display-publication-links.html pub=post%}
Funders & Partners
+
+
+-
+ {% for pass in (1..2) %}
+ {% for logo in funders %}
+
+ {% endfor %}
+ {% endfor %}
+
diff --git a/_includes/masthead.html b/_includes/masthead.html
index e14f2e4fd..5693fa8bc 100644
--- a/_includes/masthead.html
+++ b/_includes/masthead.html
@@ -25,10 +25,16 @@
%}>{{ link.tab }}
diff --git a/_includes/page__hero_particles.html b/_includes/page__hero_particles.html
index 8bc696f18..120ca013b 100644
--- a/_includes/page__hero_particles.html
+++ b/_includes/page__hero_particles.html
@@ -1,3 +1,17 @@
+{% assign research_logo_url = nil %}
+{% if page.category == "research-directions" or page.collection == "research_directions" %}
+{% case page.title %}
+{% when "Temporal Graph Learning" %}
+{% assign research_logo_url = "/assets/images/home/TGL.png" %}
+{% when "Crime & Online Markets" %}
+{% assign research_logo_url = "/assets/images/home/infrared.png" %}
+{% when "Politics & Online Media" %}
+{% assign research_logo_url = "/assets/images/home/AI4.png" %}
+{% when "Toxicity & Online Games" %}
+{% assign research_logo_url = "/assets/images/home/toxicity.png" %}
+{% endcase %}
+{% endif %}
+
{% if page.header.overlay_color or page.header.overlay_image %}
{% capture overlay_style %}
{% if page.header.overlay_image %}
@@ -17,15 +31,20 @@
{% endcapture %}
{% endif %}
-
+
+ {% unless page.url == "/" %}
+ {% endunless %}
-
+
+ {% unless research_logo_url %}
+ {% endunless %}
{% if page.title %}
+ {% if research_logo_url %}{% endif -%} {{ page.title | markdownify | remove: "
" | remove: "
" }} diff --git a/_includes/posts-highlighted-publications.html b/_includes/posts-highlighted-publications.html index 3ca032c3e..11e72935d 100644 --- a/_includes/posts-highlighted-publications.html +++ b/_includes/posts-highlighted-publications.html @@ -1,9 +1,45 @@ + -
- {%- for post in site.categories[include.taxonomy] -%}
- {%- unless post.hidden -%}
+{%- assign limit = include.limit | default: 3 -%}
+{%- assign visible_posts = site.categories[include.taxonomy] | where_exp: "post", "post.hidden != true" -%}
+{%- assign extra_count = visible_posts.size | minus: limit -%}
+
+
+
+{%- if extra_count > 0 -%}
+
+{%- endif -%}
diff --git a/_includes/research-directions.html b/_includes/research-directions.html
index 027563434..856063d39 100644
--- a/_includes/research-directions.html
+++ b/_includes/research-directions.html
@@ -5,12 +5,38 @@
{% endif %}
{% assign sorted-research-directions = research-directions | sort: "order" %}
+{% assign home-logos = "TGL.png,infrared.png,AI4.png,toxicity.png" | split: "," %}
-
+ {%- for post in visible_posts limit: limit -%}
{% include display-highlighted-publications.html %}
- {%- endunless -%}
{%- endfor -%}
-
\ No newline at end of file
+
+ {%- if extra_count > 0 -%}
+
+ {%- for post in visible_posts offset: limit -%}
+ {% include display-highlighted-publications.html %}
+ {%- endfor -%}
+
+
+
+
+
+ {%- endif -%}
+
+
+ {% if page.url == "/" %}
+
+ {% endif %}
+
+
{% for r in sorted-research-directions %}
-
+
+ {% if page.url == "/" %}
+ {% assign logo_name = home-logos[forloop.index0] %}
+ {% if logo_name %}
+
+ 
+
+ {% endif %}
+
+
+
+ {% endif %}
+
+
+
+ {% else %}
+
+
+ {% if r.one-liner %}
+ {{ r.title }}
+
+ {{ r.one-liner | markdownify }}
+
+ {% endif %}
+
+ {% for link in links %}
+
+
+ {{ link.title }}
+
+ {% endfor %}
+
+{% endif %}
diff --git a/_includes/tech-transfer.html b/_includes/tech-transfer.html
index eb3163f3d..7de44c838 100644
--- a/_includes/tech-transfer.html
+++ b/_includes/tech-transfer.html
@@ -4,12 +4,79 @@
{% assign tech_items = site.tech_transfer %}
{% endif %}
-{% assign sorted_items = tech_items | sort: "order" %}
+{% comment %} Newest paper first; entries without a paper_date go last. {% endcomment %}
+{% assign dated_items = tech_items | where_exp: "item", "item.paper_date" | sort: "paper_date" | reverse %}
+{% assign undated_items = tech_items | where_exp: "item", "item.paper_date == nil" %}
+{% assign sorted_items = dated_items | concat: undated_items %}
+
+{% if include.sortable %}
+
+
+
+{% endif %}
{% for item in sorted_items %}
-
+
+
diff --git a/_layouts/publication.html b/_layouts/publication.html
index 04488da8b..232c2ebd9 100644
--- a/_layouts/publication.html
+++ b/_layouts/publication.html
@@ -33,6 +33,15 @@
}
+ All publications
+
+
{{ content }}
+
\ No newline at end of file
+
{{ item.content | markdownify }}
+
+ {% if item.links or item.pip %}
+
{% endfor %}
-
+ {% if item.pip %}
+
+ {% endif %}
+ {% for link in item.links %}
+ {% case link.type %}
+ {% when "demo" %}{% assign icon = "fas fa-play" %}
+ {% when "dataset" %}{% assign icon = "fas fa-database" %}
+ {% when "paper" %}{% assign icon = "far fa-file-lines" %}
+ {% when "code" %}{% assign icon = "fab fa-github" %}
+ {% when "docs" %}{% assign icon = "fas fa-book" %}
+ {% when "extension" %}{% assign icon = "fab fa-chrome" %}
+ {% else %}{% assign icon = "fas fa-globe" %}
+ {% endcase %}
+
+ {{ link.label }}
+
+ {% endfor %}
+
+ {% endif %}
diff --git a/_pages/about.md b/_pages/about.md
index d5bcd8856..be4e1e2c4 100644
--- a/_pages/about.md
+++ b/_pages/about.md
@@ -18,6 +18,15 @@ row_code:
btn_class: "btn--primary"
excerpt: "We advocate for open-source code by publishing our models and datasets on GitHub, empowering researchers and developers to easily reproduce, extend, and contribute to our work through community-driven pull requests and issue discussions."
+row_datasets:
+ - image_path: /assets/images/home/huggingface.webp
+ url: https://huggingface.co/ComplexDataLab
+ alt: "Our Hugging Face page"
+ title: "Public Datasets"
+ btn_label: "Hugging Face"
+ btn_class: "btn--primary"
+ excerpt: "We release our datasets publicly on Hugging Face for benchmarking and further research."
+
row_about_us:
- image_path: /assets/images/home/misinfo_gov_outreach.webp
- image_path: /assets/images/home/group-photo-1.webp
@@ -32,6 +41,8 @@ row_about_us:
{% include feature_row id="row_code" type="right" %}
+{% include feature_row id="row_datasets" type="left" %}
+
# Join Us diff --git a/_pages/home.md b/_pages/home.md index 577191d52..a8d55417c 100644 --- a/_pages/home.md +++ b/_pages/home.md @@ -7,6 +7,27 @@ header: overlay_css: "background-size: contain; background-repeat: no-repeat;" excerpt: "A research group focusing on developing techniques for analyzing complex data from online societies, with applications to enhance the health and safety of online spaces." +# Homepage funders strip: logos are collected from the research directions' `logos`; +# `exclude` drops some (by image path) and `extra` adds funders not listed on a research page. +funders: + exclude: + - /assets/images/logo/GOOGLE.png + - /assets/images/logo/SAMSUNG.png + extra: + - image: /assets/images/logo/FLI.png + name: Future of Life Institute + +research_links: + - title: "Tools & Resources" + url: /tools-and-resources/ + icon: "fas fa-toolbox" + - title: "Publications" + url: /publications/ + icon: "fas fa-book-open" + - title: "STAMINA Reading Group" + url: /stamina/ + icon: "fas fa-users" + row_research: - image_path: /assets/images/home/group-photo-2.webp url: /publications @@ -20,4 +41,8 @@ row_research: {% include research-directions.html category="research-directions"%} +{% include research-links.html %} + {% include feature_row id="row_research" type="left" %} + +{% include funders-carousel.html %} diff --git a/_pages/people.md b/_pages/people.md index d1485671f..e4d9e777f 100644 --- a/_pages/people.md +++ b/_pages/people.md @@ -7,7 +7,7 @@ header: overlay_css: "background-size: contain; background-repeat: no-repeat;" --- -## Faculty & Staff & Postdoc +## Faculty, Staff & Postdoc {% include card-authors-with-role.html authors=site.data.authors role="Faculty" render_current_role=true %} {% include card-authors-with-role.html authors=site.data.authors role="Staff" render_current_role=true %} @@ -15,49 +15,31 @@ header: ## Students -### PhD +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="PhD" heading="PhD" heading_id="phd" render_current_role=true %} -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="PhD" render_current_role=true %} +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Master" heading="Masters" heading_id="masters" render_current_role=true %} -### Masters +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Undergraduate" heading="Undergraduate" heading_id="undergraduate" render_current_role=true %} -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Master" render_current_role=true %} - -### Undergraduate - -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Undergraduate" render_current_role=true %} - -### Intern - -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Intern" render_current_role=true %} +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Intern" heading="Intern" heading_id="intern" render_current_role=true %} --- # Past Members -## Faculty & Staff & Postdoc +## Faculty, Staff & Postdocs -{% include card-authors-with-role.html authors=site.data.authors role="Faculty" render_current_role=false %} -{% include card-authors-with-role.html authors=site.data.authors role="Staff" render_current_role=false %} -{% include card-authors-with-role.html authors=site.data.authors role="Postdoc" render_current_role=false %} +{% include card-authors-with-role.html authors=site.data.authors roles="Faculty,Staff,Postdoc" render_current_role=false %} ## Alumni -### PhD - -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="PhD" render_current_role=false %} - -### Masters - -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Master" render_current_role=false %} - -### Undergraduate +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="PhD" heading="PhD" heading_id="phd-past" render_current_role=false %} -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Undergraduate" render_current_role=false %} +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Master" heading="Masters" heading_id="masters-past" render_current_role=false %} -### Interns +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Undergraduate" heading="Undergraduate" heading_id="undergraduate-past" render_current_role=false %} -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Intern" render_current_role=false %} +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Intern" heading="Interns" heading_id="interns-past" render_current_role=false %}
diff --git a/_pages/publications.md b/_pages/publications.md index c8af1c166..b1c68065f 100644 --- a/_pages/publications.md +++ b/_pages/publications.md @@ -8,6 +8,7 @@ excerpt: "" classes: - no-sidebar - wide + - publications-page permalink: /publications/ taxonomy: Publications show_view_by_venue: true diff --git a/_pages/tools-and-resources.md b/_pages/tools-and-resources.md index 4ddd8dac8..b469f0020 100644 --- a/_pages/tools-and-resources.md +++ b/_pages/tools-and-resources.md @@ -10,4 +10,4 @@ excerpt: "Showcasing our tools and resources that bridge research innovations wi --- -{% include tech-transfer.html %} +{% include tech-transfer.html sortable=true %} diff --git a/_posts/news/2025-06-29-TGL_2025.md b/_posts/news/2025-06-29-TGL_2025.md index 923b87a26..eb85d1154 100644 --- a/_posts/news/2025-06-29-TGL_2025.md +++ b/_posts/news/2025-06-29-TGL_2025.md @@ -1,6 +1,6 @@ --- title: "TGL'25" -venue: Temporal Graph Learning Workshop (Third Edition) @ KDD 2025! +venue: Temporal Graph Learning Workshop (Third Edition) @ KDD 2025 names: Farimah Poursafaei, Shenyang Huang, Julia Gastinger, Daniele Zambon, Andrea Cini, Jacob Chmura, Reihaneh Rabbany, Michael Bronstein author: Farimah Poursafaei categories: News diff --git a/_posts/news/2025-07-01-Social_Sim_2025.md b/_posts/news/2025-07-01-Social_Sim_2025.md index 06823b2e5..86f1a3640 100644 --- a/_posts/news/2025-07-01-Social_Sim_2025.md +++ b/_posts/news/2025-07-01-Social_Sim_2025.md @@ -1,6 +1,6 @@ --- title: "Social Sim'25" -venue: Social Simulation with LLMs has been accepted at COLM 2025! +venue: Social Simulation with LLMs has been accepted at COLM 2025 names: Zachary Yang, Kellin Pelrine, Maximillian Puelma Touzel, Dan Zhao, Jean-Francois Godbout, Petter Törnberg, Reihaneh Rabbany, William Cunningham author: Zachary Yang categories: News diff --git a/_posts/news/2025-07-09-FIIR_2025.md b/_posts/news/2025-07-09-FIIR_2025.md index cf056201d..447afeace 100644 --- a/_posts/news/2025-07-09-FIIR_2025.md +++ b/_posts/news/2025-07-09-FIIR_2025.md @@ -1,6 +1,6 @@ --- title: "FIIR'25" -venue: Future of Information Integrity Research will be co-located with ICDM'25! +venue: Future of Information Integrity Research will be co-located with ICDM'25 names: Reihaneh Rabbany, Kai Shu, Jean-Francois Godbout, Taylor Lynn Curtis, Bijean Ghafouri, Kellin Pelrine, Maximilian Puelma Touzel author: Reihaneh Rabbany categories: News diff --git a/_posts/news/2026-10-09-Social_Sim_2026.md b/_posts/news/2026-10-09-Social_Sim_2026.md new file mode 100644 index 000000000..fdb632e0e --- /dev/null +++ b/_posts/news/2026-10-09-Social_Sim_2026.md @@ -0,0 +1,11 @@ +--- +title: "Social Sim'26" +venue: The 2nd Workshop on Social Simulation with LLMs at COLM 2026 +names: Zachary Yang, Maximilian Puelma Touzel, Xuhui Zhou, Yunze Xiao, Lynnette Hui Xian Ng, Jean-Francois Godbout, Reihaneh Rabbany, Maarten Sap +author: Zachary Yang +categories: News +thumbnail: "assets/images/news/social_sim.jfif" +--- + + +We’re pleased to announce Social Sim’26, the 2nd Workshop on Social Simulation with LLMs, taking place at COLM 2026 in San Francisco on October 9. This year’s theme is *Fidelity in Applications*. The workshop also hosts the *Many Worlds* shared task on evaluating LLM social simulations, with [code](https://github.com/sandbox-social/socsim26_sharedtask) and [data](https://huggingface.co/datasets/ComplexDataLab/socsim26-sharedtask) openly available. For more information, please refer to the [workshop website](https://sites.google.com/view/social-sims-with-llms/). diff --git a/_posts/papers/2024-11-07-2411.05060.md b/_posts/papers/2024-11-07-2411.05060.md deleted file mode 100644 index 58383d5cf..000000000 --- a/_posts/papers/2024-11-07-2411.05060.md +++ /dev/null @@ -1,28 +0,0 @@ ---- -title: A Guide to Misinformation Detection Datasets -venue: arXiv.org -names: Camille Thibault, Gabrielle Peloquin-Skulski, Jacob-Junqi Tian, Florence Laflamme, - Yuxiang Guan, Reihaneh Rabbany, J. Godbout, Kellin Pelrine -tags: -- arXiv.org -link: https://arxiv.org/abs/2411.05060 -webpage: https://misinfo-datasets.complexdatalab.com/ -code: https://github.com/ComplexData-MILA/misinfo-datasets -author: Jacob-Junqi Tian -categories: -- Publications -- poli-sci -excerpt: Data is a barrier to reliable misinformation detection solutions. To address this, we curated the largest collection of (mis)information datasets in the literature, totaling 75. From these, we evaluated the quality of all of the 36 datasets that consist of statements or claims. We assessed these datasets to identify those with solid foundations for empirical work and those with flaws that could result in misleading results, such as insufficient label quality, spurious correlations, or political bias. We further show that regardless of label quality, categorical labels may no longer give an accurate evaluation of detection model performance, and discuss ways to solve this. Overall, this guide aims to provide a roadmap for obtaining higher quality data and conducting more effective evaluations in this domain. - - ---- - -*{{ page.names }}* - -**{{ page.venue }}** - -{% include display-publication-links.html pub=page %} - -## Abstract - -Misinformation is a complex societal issue, and mitigating solutions are difficult to create due to data deficiencies. To address this problem, we have curated the largest collection of (mis)information datasets in the literature, totaling 75. From these, we evaluated the quality of all of the 36 datasets that consist of statements or claims. We assess these datasets to identify those with solid foundations for empirical work and those with flaws that could result in misleading and non-generalizable results, such as insufficient label quality, spurious correlations, or political bias. We further provide state-of-the-art baselines on all these datasets, but show that regardless of label quality, categorical labels may no longer give an accurate evaluation of detection model performance. We discuss alternatives to mitigate this problem. Overall, this guide aims to provide a roadmap for obtaining higher quality data and conducting more effective evaluations, ultimately improving research in misinformation detection. All datasets and other artifacts are available at https://misinfo-datasets.complexdatalab.com/. \ No newline at end of file diff --git a/_posts/papers/2025-02-25-10.1609-aaai.v39i27.35005.md b/_posts/papers/2025-02-25-10.1609-aaai.v39i27.35005.md new file mode 100644 index 000000000..5e37b0e50 --- /dev/null +++ b/_posts/papers/2025-02-25-10.1609-aaai.v39i27.35005.md @@ -0,0 +1,24 @@ +--- +title: "Enhancing Privacy in the Early Detection of Sexual Predators Through Federated Learning and Differential Privacy" +venue: Proceedings of the Thirty-ninth AAAI Conference on Artificial Intelligence, AI for Social Impact (AISI) track +names: Khaoula Chehbouni, Martine De Cock, Gilles Caporossi, Afaf Taik, Reihaneh Rabbany, Golnoosh Farnadi +tags: +- AAAI +link: https://doi.org/10.1609/aaai.v39i27.35005 +author: Khaoula Chehbouni +categories: +- Publications +- online-crime +excerpt: Existing approaches to detecting online grooming monitor private conversations through centrally trained models. We implement a privacy-preserving pipeline for the early detection of sexual predators using federated learning and differential privacy, and show on real-world data that privacy and utility can coexist with only a slight reduction in utility. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +The increased screen time and isolation caused by the COVID-19 pandemic have led to a significant surge in cases of online grooming, which is the use of strategies by predators to lure children into sexual exploitation. Previous efforts to detect grooming in industry and academia have involved accessing and monitoring private conversations through centrally-trained models or sending private conversations to a global server. In this work, we implement a privacy-preserving pipeline for the early detection of sexual predators. We leverage federated learning and differential privacy in order to create safer online spaces for children while respecting their privacy. We investigate various privacy-preserving implementations and discuss their benefits and shortcomings. Our extensive evaluation using real-world data proves that privacy and utility can coexist with only a slight reduction in utility. diff --git a/_posts/papers/2025-08-03-10.1145-3711896.3737437.md b/_posts/papers/2025-08-03-10.1145-3711896.3737437.md new file mode 100644 index 000000000..6cfd7d21e --- /dev/null +++ b/_posts/papers/2025-08-03-10.1145-3711896.3737437.md @@ -0,0 +1,29 @@ +--- +title: "A Guide to Misinformation Detection Data and Evaluation" +venue: Proceedings of the 31st SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), Datasets and Benchmarks Track +names: Camille Thibault, Jacob-Junqi Tian, Gabrielle Péloquin-Skulski, Taylor Lynn Curtis, James Zhou, + Florence Laflamme, Yuxiang Guan, Reihaneh Rabbany, Jean-François Godbout, Kellin Pelrine +tags: +- KDD +link: https://dl.acm.org/doi/10.1145/3711896.3737437 +webpage: https://misinfo-datasets.complexdatalab.com/ +code: https://github.com/ComplexData-MILA/misinfo-datasets +award: Best Datasets & Benchmarks Track Paper, Runner-Up +award_url: https://kdd2025.kdd.org/awards/ +author: Camille Thibault +categories: +- Publications +- poli-sci +excerpt: We curated the largest collection of (mis)information datasets in the literature, totaling 75, and evaluated the quality of the 45 claim- and paragraph-based ones, finding that spurious correlations and ambiguous, unverifiable examples affect most of them. We provide state-of-the-art baselines, show that categorical labels may no longer accurately evaluate detection models, and propose Evaluation Quality Assurance (EQA) to guide the field toward better grounded evaluations. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Misinformation is a complex societal issue, and mitigating solutions are difficult to create due to data deficiencies. To address this, we have curated the largest collection of (mis)information datasets in the literature, totaling 75. From these, we evaluated the quality of 36 datasets that consist of statements or claims, as well as the 9 datasets that consist of data in purely paragraph form. We assess these datasets to identify those with solid foundations for empirical work and those with flaws that could result in misleading and non-generalizable results, such as spurious correlations, or examples that are ambiguous or otherwise impossible to assess for veracity. We find the latter issue is particularly severe and affects most datasets in the literature. We further provide state-of-the-art baselines on all these datasets, but show that regardless of label quality, categorical labels may no longer give an accurate evaluation of detection model performance. Finally, we propose and highlight Evaluation Quality Assurance (EQA) as a tool to guide the field toward systemic solutions rather than inadvertently propagating issues in evaluation. Overall, this guide aims to provide a roadmap for higher quality data and better grounded evaluations, ultimately improving research in misinformation detection. All datasets and other artifacts are available at https://misinfo-datasets.complexdatalab.com/. diff --git a/_posts/papers/2025-10-15-10.1609-aies.v8i2.36677.md b/_posts/papers/2025-10-15-10.1609-aies.v8i2.36677.md new file mode 100644 index 000000000..e441811b0 --- /dev/null +++ b/_posts/papers/2025-10-15-10.1609-aies.v8i2.36677.md @@ -0,0 +1,24 @@ +--- +title: "Reframing AI-for-Good: Radical Questioning in AI for Human Trafficking Interventions" +venue: Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society (AIES) +names: Pratheeksha Nair, Gabriel Lefebvre, Maryam Molamohammadi, Sophia Garrel, Reihaneh Rabbany +tags: +- AIES +link: https://doi.org/10.1609/aies.v8i2.36677 +author: Pratheeksha Nair +categories: +- Publications +- online-crime +excerpt: We introduce Radical Questioning, a five-step, pre-design ethics framework for asking whether AI should be applied to a complex social problem, not just how. Developed with survivor-led organizations and piloted in human trafficking, it shifted our design away from automated detection tools and toward survivor-controlled, empowerment-based technologies. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +This paper introduces Radical Questioning (RQ), a structured, pre-design ethics framework developed to assess whether artificial intelligence (AI) should be applied to complex social problems rather than merely how. While much of responsible AI development focuses on aligning systems with principles such as fairness, transparency, and accountability, it often begins after the decision to build has already been made, implicitly treating the deployment of AI as a given rather than a question in itself. In domains such as human trafficking, marked by contested definitions, systemic injustice, and deep stakeholder asymmetries, such assumptions can obscure foundational ethical concerns. RQ offers an upstream, deliberative process for surfacing these concerns before design begins. Drawing from critical theory, participatory ethics, and relational responsibility, RQ formalizes a five-step framework to interrogate problem framings, confront techno-solutionist tendencies, and reflect on the moral legitimacy of intervention. Developed through interdisciplinary collaboration and engagement with survivor-led organizations, RQ was piloted in the domain of human trafficking (HT) which is a particularly high-stakes and ethically entangled application area. Its use led to a fundamental design shift: away from automated detection tools and toward survivor-controlled, empowerment-based technologies. We argue that RQ's novelty lies in both its temporal position, i.e, prior to technical design, and its orientation toward domains where harm is structural and ethical clarity cannot be achieved through one-size-fits-all solutions. RQ thus addresses a critical gap between abstract principles of responsible AI and the lived ethical demands of real-world deployment. diff --git a/_posts/papers/2025-12-02-2406.10426.md b/_posts/papers/2025-12-02-2406.10426.md new file mode 100644 index 000000000..860148320 --- /dev/null +++ b/_posts/papers/2025-12-02-2406.10426.md @@ -0,0 +1,26 @@ +--- +title: "MiNT: Multi-Network Transfer Benchmark for Temporal Graph Learning" +venue: Proceedings of the Thirty-ninth Conference on Neural Information Processing Systems (NeurIPS), Datasets and Benchmarks Track +names: Kiarash Shamsi, Tran Gia Bao Ngo, Razieh Shirzadkhani, Shenyang Huang, Farimah Poursafaei, Poupak Azad, + Reihaneh Rabbany, Baris Coskunuzer, Guillaume Rabusseau, Cuneyt Gurcan Akcora +tags: +- NeurIPS +link: https://papers.neurips.cc/paper_files/paper/2025/hash/c5548168cb7324f714365a971dfe76d1-Abstract-Datasets_and_Benchmarks_Track.html +code: https://github.com/benjaminnNgo/ScalingTGNs +author: Shenyang Huang +categories: +- Publications +- TGL +excerpt: We introduce a benchmark of 84 real-world temporal transaction networks and MiNT, a pre-training framework that captures transferable temporal dynamics across networks. Trained on up to 64 networks and evaluated on 20 unseen ones, MiNT consistently outperforms individually trained models, revealing scaling trends in temporal graph learning and laying the groundwork for temporal graph foundation models. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Temporal Graph Learning (TGL) aims to discover patterns in evolving networks or temporal graphs and leverage these patterns to predict future interactions. However, most existing research focuses on learning from a single network in isolation, leaving the challenges of within-domain and cross-domain generalization largely unaddressed. In this study, we introduce a new benchmark of 84 real-world temporal transaction networks and propose Temporal Multi-network Transfer (MiNT), a pre-training framework designed to capture transferable temporal dynamics across diverse networks. We train MiNT models on up to 64 transaction networks and evaluate their generalization ability on 20 held-out, unseen networks. Our results show that MiNT consistently outperforms individually trained models, revealing a strong relation between the number of pre-training networks and transfer performance. These findings highlight scaling trends in temporal graph learning and underscore the importance of network diversity in improving generalization. This work establishes the first large-scale benchmark for studying transferability in TGL and lays the groundwork for developing Temporal Graph Foundation Models. diff --git a/_posts/papers/2026-04-13-2512.13915.md b/_posts/papers/2026-04-13-2512.13915.md new file mode 100644 index 000000000..2401800ff --- /dev/null +++ b/_posts/papers/2026-04-13-2512.13915.md @@ -0,0 +1,24 @@ +--- +title: "Deepfakes in the 2025 Canadian Election: Prevalence, Partisanship, and Platform Dynamics" +venue: Proceedings of The Web Conference (TheWebConf), Short Paper Track +names: Victor Livernoche, Andreea Musulan, Zachary Yang, Jean-François Godbout, Reihaneh Rabbany +tags: +- TheWebConf +link: https://arxiv.org/abs/2512.13915 +author: Victor Livernoche +categories: +- Publications +- poli-sci +excerpt: Analyzing 187,778 posts from X, Bluesky and Reddit during the 2025 Canadian federal election, we find that 5.86% of election-related images were deepfakes, shared more often by right-leaning accounts. Most were benign or non-political and harmful ones drew little attention, but realistic fabricated images drew higher engagement. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Concerns about AI-generated political content are growing, yet there is limited empirical evidence on how deepfakes actually appear and circulate across social platforms during major events in democratic countries. In this study, we present one of the first in-depth analyses of how these realistic synthetic media shape the political landscape online, focusing specifically on the 2025 Canadian federal election. By analyzing 187,778 posts from X, Bluesky, and Reddit with a high-accuracy detection framework trained on a diverse set of modern generative models, we find that 5.86% of election-related images were deepfakes. Right-leaning accounts shared them more frequently, with 8.66% of their posted images flagged compared to 4.42% for left-leaning users, often with defamatory or conspiratorial intent. Yet, most detected deepfakes were benign or non-political, and harmful ones drew little attention, accounting for only 0.12% of all views on X. Overall, deepfakes were present in the election conversation, but their reach was modest, and realistic fabricated images, although less common, drew higher engagement, highlighting growing concerns about their potential misuse. diff --git a/_posts/papers/2026-04-23-2510.07586.md b/_posts/papers/2026-04-23-2510.07586.md new file mode 100644 index 000000000..b83ce3975 --- /dev/null +++ b/_posts/papers/2026-04-23-2510.07586.md @@ -0,0 +1,27 @@ +--- +title: "TGM: A Modular and Efficient Library for Machine Learning on Temporal Graphs" +venue: Proceedings of the Fourteenth International Conference on Learning Representations (ICLR) +names: Jacob Chmura, Shenyang Huang, Tran Gia Bao Ngo, Ali Parviz, Farimah Poursafaei, Jure Leskovec, + Michael M. Bronstein, Guillaume Rabusseau, Matthias Fey, Reihaneh Rabbany +tags: +- ICLR +link: https://arxiv.org/abs/2510.07586 +code: https://github.com/tgm-team/tgm +webpage: https://tgm.readthedocs.io +author: Shenyang Huang +categories: +- Publications +- TGL +excerpt: We introduce TGM, a research-oriented library for machine learning on temporal graphs and the first to unify continuous- and discrete-time dynamic graph methods. It supports dynamic node features, time-granularity conversions and link-, node- and graph-level tasks, with an average 7.8x speedup over DyGLib and 175x on graph discretization. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Well-designed open-source software drives progress in Machine Learning (ML) research. While static graph ML enjoys mature frameworks like PyTorch Geometric and DGL, ML for temporal graphs (TG), networks that evolve over time, lacks comparable infrastructure. Existing TG libraries are often tailored to specific architectures, hindering support for diverse models in this rapidly evolving field. Additionally, the divide between continuous- and discrete-time dynamic graph methods (CTDG and DTDG) limits direct comparisons and idea transfer. To address these gaps, we introduce Temporal Graph Modelling (TGM), a research-oriented library for ML on temporal graphs, the first to unify CTDG and DTDG approaches. TGM offers first-class support for dynamic node features, time-granularity conversions, and native handling of link-, node-, and graph-level tasks. Empirically, TGM achieves an average 7.8x speedup across multiple models, datasets, and tasks compared to the widely used DyGLib, and an average 175x speedup on graph discretization relative to available implementations. Beyond efficiency, we show in our experiments how TGM unlocks entirely new research possibilities by enabling dynamic graph property prediction and time-driven training paradigms, opening the door to questions previously impractical to study. TGM is available at https://github.com/tgm-team/tgm. diff --git a/_posts/papers/2026-04-23-2511.07332.md b/_posts/papers/2026-04-23-2511.07332.md new file mode 100644 index 000000000..f756812a8 --- /dev/null +++ b/_posts/papers/2026-04-23-2511.07332.md @@ -0,0 +1,24 @@ +--- +title: "Grounding Computer Use Agents on Human Demonstrations" +venue: Proceedings of the Fourteenth International Conference on Learning Representations (ICLR) +names: Aarash Feizi, Shravan Nayak, Xiangru Jian, Kevin Qinghong Lin, Kaixin Li, Rabiul Awal, Xing Han Lù, + Johan Obando-Ceron, Juan A. Rodriguez, Nicolas Chapados, David Vazquez, Adriana Romero-Soriano, + Reihaneh Rabbany, Perouz Taslakian, Christopher Pal, Spandana Gella, Sai Rajeswar +tags: +- ICLR +link: https://arxiv.org/abs/2511.07332 +author: Aarash Feizi +categories: Publications +excerpt: We introduce GroundCUA, a large-scale desktop grounding dataset built from expert human demonstrations, covering 87 applications and over 3.56M human-verified annotations. The GroundNext models trained on it achieve state-of-the-art grounding across five benchmarks with less than one-tenth the training data of prior work. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Building reliable computer-use agents requires grounding: accurately connecting natural language instructions to the correct on-screen elements. While large datasets exist for web and mobile interactions, high-quality resources for desktop environments are limited. To address this gap, we introduce GroundCUA, a large-scale desktop grounding dataset built from expert human demonstrations. It covers 87 applications across 12 categories and includes 56K screenshots, with every on-screen element carefully annotated for a total of over 3.56M human-verified annotations. From these demonstrations, we generate diverse instructions that capture a wide range of real-world tasks, providing high-quality data for model training. Using GroundCUA, we develop the GroundNext family of models that map instructions to their target UI elements. At both 3B and 7B scales, GroundNext achieves state-of-the-art results across five benchmarks using supervised fine-tuning, while requiring less than one-tenth the training data of prior work. Reinforcement learning post-training further improves performance, and when evaluated in an agentic setting on the OSWorld benchmark using o3 as planner, GroundNext attains comparable or superior results to models trained with substantially more data. These results demonstrate the critical role of high-quality, expert-driven datasets in advancing general-purpose computer-use agents. diff --git a/_posts/papers/2026-05-25-pmlr-v318-yang26a.md b/_posts/papers/2026-05-25-pmlr-v318-yang26a.md new file mode 100644 index 000000000..2497cb889 --- /dev/null +++ b/_posts/papers/2026-05-25-pmlr-v318-yang26a.md @@ -0,0 +1,24 @@ +--- +title: "ToxiSight: Leveraging Moderator Expertise Through Behavioral Measurement in Gaming Toxicity Annotation" +venue: Proceedings of the Canadian Conference on Artificial Intelligence (Canadian AI) +names: Zachary Yang, Vicki Chen, Domenico Tullo, Reihaneh Rabbany +tags: +- Canadian AI +link: https://proceedings.mlr.press/v318/yang26a.html +author: Zachary Yang +categories: +- Publications +- online-toxicity +excerpt: We present ToxiSight, an annotation platform that treats moderators' hesitation, revision and disagreement as signals of genuinely ambiguous content and failing guidelines, rather than resolving them into consensus. Deployed with 10 professional moderators across 60,000 lines of gaming chat, it exposed systematic category failures invisible to inter-annotator metrics and raised completion rates from 60% to 95%. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Content moderation systems commonly treat human annotators as interchangeable label sources, resolving disagreements through majority voting or expert arbitration. We present ToxiSight, an annotation platform that reframes this assumption: rather than extracting consensus, the system supports moderator reasoning by treating hesitation, revision, and disagreement as signals revealing where content is genuinely ambiguous and where taxonomic guidelines fail. ToxiSight integrates gaming-specific contextual widgets with behavioral telemetry, capturing the cognitive processes underlying toxicity validation decisions. Through deployment with 10 professional moderators across 60,000 lines of gaming chat, we demonstrate that behavioral patterns expose systematic category failures invisible to traditional inter-annotator metrics. The Controversial category shows 72% revision rates with fast processing times, indicating immediate recognition of definitional breakdown, while Threats (Life-Threatening) exhibits 75% revisions with slow processing, signaling genuine interpretive complexity. Completion rates improved from 60% to 95%, and moderators reported reduced decision stress when permitted to express uncertainty. This case study demonstrates that trustworthy toxicity detection requires annotation systems designed around the irreducible complexity of human judgment, not against it. diff --git a/_posts/papers/2026-06-01-ssrn.6867358.md b/_posts/papers/2026-06-01-ssrn.6867358.md new file mode 100644 index 000000000..fa3c06400 --- /dev/null +++ b/_posts/papers/2026-06-01-ssrn.6867358.md @@ -0,0 +1,28 @@ +--- +title: "AI Epistemic Risks: Emerging Mechanisms & Evidence" +venue: SSRN +names: Stephen Casper, Jonathan Stray, Anna Gausen, Cameron Jones, Brian Christian, Bálint Gyevnár, + Siao Si Looi, Joshua Levy, Elizabeth Seger, Hannah Kirk, Kobi Hackenburg, Matt Kowal, Michelle Malonza, + Dan Zhao, Zhonghao He, Luke Hewitt, Hause Lin, Maarten Sap, Dylan Hadfield-Menell, Atoosa Kasirzadeh, + Gordon Pennycook, Jean-François Godbout, Kellin Pelrine, Reihaneh Rabbany, Natasha Jaques, Jasmine Li, + Yoshua Bengio, David Rand, Thomas Costello, Mick Yang +tags: +- SSRN +link: https://ssrn.com/abstract=6867358 +author: Jean-Francois Godbout +categories: +- Publications +- poli-sci +excerpt: AI poses risks to humanity's collective capacity to form accurate beliefs, reason well and maintain a healthy information environment. We identify three mechanisms of systemic epistemic decline (persuasion and manipulation, cognitive offloading, and feedback loops), record the emerging empirical evidence for each, and analyze how they amplify one another. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Advances in artificial intelligence pose risks to humanity's collective capacity to form accurate beliefs, reason well, and maintain a healthy information environment: risks we term epistemic risks. These risks arise from AI's integration into the infrastructure through which individuals and societies think, form beliefs, and make sense of the world together. We highlight three primary mechanisms by which AI can cause systemic epistemic decline: persuasion and manipulation, cognitive offloading, and feedback loops. AI can persuade and manipulate through misuse for economic or political manipulation, inciting crime or radicalization, and escalating conflict, and can also cause unintentional harms such as sycophancy and mental health risks. We record the emerging empirical evidence for each mechanism and analyze how they likely amplify one another, creating compounding systemic risk. The evidence suggests that AI could be an unprecedented lever for improving epistemics, but this will not happen by default. diff --git a/_posts/papers/2026-07-06-icml2026-validation-gap.md b/_posts/papers/2026-07-06-icml2026-validation-gap.md new file mode 100644 index 000000000..532ad0686 --- /dev/null +++ b/_posts/papers/2026-07-06-icml2026-validation-gap.md @@ -0,0 +1,18 @@ +--- +title: "Position: Time to Close The Validation Gap in LLM Social Simulations" +venue: Proceedings of the Forty-third International Conference on Machine Learning (ICML), Position Paper Track +names: Maximilian Puelma Touzel, Sneheel Sarangi, Aurélien Bück-Kaeffer, Zachary Yang, Jean-François Godbout, Reihaneh Rabbany +tags: +- ICML +link: https://openreview.net/forum?id=LpbxLBcOBf +author: Maximilian Puelma Touzel +categories: Publications +excerpt: LLM-based social simulations are rapidly proliferating across policy analysis, epidemiology and computational social science, yet the field lacks consensus on how to validate them. We argue for shifting from expansion to consolidation, through shared benchmarks, open data and reproducible evaluation grounded in social science, and outline a concrete research program toward that goal. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} diff --git a/_posts/papers/2026-08-09-10.1145-3770855.3817476.md b/_posts/papers/2026-08-09-10.1145-3770855.3817476.md new file mode 100644 index 000000000..a3460c84c --- /dev/null +++ b/_posts/papers/2026-08-09-10.1145-3770855.3817476.md @@ -0,0 +1,25 @@ +--- +title: "CrediBench: Web-Scale Network Data for Credibility Prediction" +venue: Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2 +names: Emma Kondrup, Sebastian Sabry, Hussein Abdallah, Zachary Yang, Jiaqi Xiong, Kellin Pelrine, + James Zhou, Zhijin Guo, Michael M Bronstein, Jean-François Godbout, Reihaneh Rabbany, Shenyang Huang +tags: +- KDD +link: https://dl.acm.org/doi/pdf/10.1145/3770855.3817476 +author: Emma Kondrup +categories: +- Publications +- poli-sci +excerpt: We present CrediBench, a web-scale dataset of temporal web graphs built from Common Crawl, with over 40 million nodes, their webpage content and over 1 billion hyperlink edges per monthly snapshot, plus credibility labels for 662,575 web domains. Combining graph, text and time signals substantially improves web domain credibility prediction over existing baselines. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Automatically assessing the credibility of online sources presents an invaluable tool for navigating today's information ecosystem. However, existing approaches either depend on scarce and costly human annotations, or focus exclusively on assessments at the level of individual claims. Misinformation often spreads via interlinked web domains, whose connections evolve over time. Focusing on claims alone ignores these structural and temporal credibility signals evident in the changing web topology. Existing datasets fail to capture these central modalities in web domain credibility prediction: namely, internet topology, temporality and text (webpage) content. To address this gap, we present CrediBench, a dataset containing eight months of web graph data; of which we analyze the three months surrounding the 2024 U.S. federal elections, a time of heightened misinformation propagation online. Each monthly snapshot contains over 40 million nodes, their scraped webpage content, and over 1 billion hyperlink edges. CrediBench supports credibility prediction as both a regression (continuous credibility score) and a binary classification task (credible or not). For classification, we curate a novel binary label set containing 662,575 web domains labelled for boolean credibility, spanning four areas (misinformation, crowd-sourced, malware and phishing). Our empirical experiments support that all task modalities (graph, text and time) contribute significantly to achieving the best performance. In particular, our multi-modal regression model trained on CrediBench outperforms other configurations and existing baselines, decreasing Mean Average Error from 0.162 to 0.107 on the regression task, while the multi-modal classifier improves accuracy from 56% to 85% on the classification one. CrediBench, our proposed web-scale multi-modal dataset, is available on Huggingface for future research. diff --git a/_posts/papers/2026-08-15-rlc2026-dart.md b/_posts/papers/2026-08-15-rlc2026-dart.md new file mode 100644 index 000000000..b73196815 --- /dev/null +++ b/_posts/papers/2026-08-15-rlc2026-dart.md @@ -0,0 +1,22 @@ +--- +title: "DART: Dual Adaptive Residual Tracking for Low-Bias Advantage Estimation and Credit Assignment in AI Agents" +venue: Proceedings of the Reinforcement Learning Conference (RLC) +names: Shahrad Mohammadzadeh, Amir-massoud Farahmand, Reihaneh Rabbany, Doina Precup +tags: +- RLC +link: https://rlj.cs.umass.edu/2026/papers/Paper74.html +author: Shahrad Mohammadzadeh +categories: Publications +excerpt: We show that the bottleneck of Generalized Advantage Estimation is dissemination bias, which can be corrected without discarding the critic. DART pairs a stable base critic with a residual critic that tracks a Monte Carlo correction, provably lowering steady-state error and matching or outperforming PPO across dm_control, especially on sparse and long-horizon tasks. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Proximal Policy Optimization (PPO) is the primary workhorse for complex continuous control and Reinforcement Learning from Human Feedback (RLHF). However, the reliability of Generalized Advantage Estimation (GAE) has recently been called into question: critic-free methods like RLOO and GRPO have raised the question of whether value-induced bias is doing more harm than the variance reduction it provides. We show this is the wrong dichotomy. The bottleneck is the dissemination bias of GAE, where value approximation error propagates into advantage estimates and misleads the policy gradient, a bias that can be corrected without discarding the critic. We propose Dual Adaptive Residual Tracking (DART), a versatile two-critic scheme that decouples the bias–variance trade-off. DART pairs a stable base critic learning from low-variance bootstrapped targets with a residual critic that tracks a high-fidelity Monte Carlo correction via a frozen-target mechanism. Theoretically, under linear function approximation assumptions, we prove that DART achieves a strictly smaller steady-state mean-squared error (MSE) upper bound than single-critic methods in high-variance regimes. Empirically, across all dm_control continuous control environments, DART matches or outperforms PPO in aggregate, with clear advantages in sparse and long-horizon tasks. We further provide initial evidence of DART's effectiveness in LLM fine-tuning. diff --git a/_posts/papers/2026-09-08-10.5753-sbbd_estendido.2026.249933.md b/_posts/papers/2026-09-08-10.5753-sbbd_estendido.2026.249933.md new file mode 100644 index 000000000..d622067a3 --- /dev/null +++ b/_posts/papers/2026-09-08-10.5753-sbbd_estendido.2026.249933.md @@ -0,0 +1,25 @@ +--- +title: "SynTraffic: A synthetic dataset for spatiotemporal human trafficking detection in heterogeneous graphs" +venue: Simpósio Brasileiro de Banco de Dados (SBBD) +names: Joao Lopes, Ana Nascimento, Pratheeksha Nair, Catalina Vajiac, Andreas Olligschlaeger, Reihaneh Rabbany, + Christos Faloutsos, Mirela Cazzolato +tags: +- SBBD +link: https://doi.org/10.5753/sbbd_estendido.2026.249933 +author: Pratheeksha Nair +categories: +- Publications +- online-crime +excerpt: Human trafficking advertisement data is sensitive and hard to share, which limits research on automated detection. We present SynTraffic, a privacy-safe synthetic dataset and reproducible generation methodology that statistically reproduces a real corpus' posting volume, spatial mobility, temporal dynamics and metadata-sharing structure. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Online escort platforms have become a common vector for human trafficking (HT) networks to advertise victims alongside legitimate listings, making manual screening infeasible. Progress on automated analysis is limited by restricted data access: HT advertisement corpora carry personal, potentially victim-identifying information, imposing ethical and legal constraints on sharing and reproducibility. This work addresses that gap with a reproducible methodology for generating synthetic HT advertisement data that is statistically representative of a real corpus, without exposing sensitive content. The proposed dataset SynTraffic reproduces posting volume, spatial mobility, temporal dynamics, and metadata-sharing structure, including the size distribution of identity clusters formed by shared contact metadata. We statistically validate SynTraffic testing distributional fidelity and cluster-size fidelity. SynTraffic is openly available, as well as the generation methodology, presenting a privacy-safe substitute for the real corpus. diff --git a/_posts/papers/2026-11-01-2605.00513.md b/_posts/papers/2026-11-01-2605.00513.md new file mode 100644 index 000000000..90807ac86 --- /dev/null +++ b/_posts/papers/2026-11-01-2605.00513.md @@ -0,0 +1,22 @@ +--- +title: "ControBench: An Interaction-Aware Benchmark for Controversial Discourse Analysis on Social Networks" +venue: The IEEE International Conference on Data Mining (ICDM) +names: Thanh Thuy Ta, Jiaqi Zhu, Xuan Liu, Shang Lin, Reihaneh Rabbany, Guillaume Rabusseau, Lihui Chen, Yilun Zheng, Sitao Luan +tags: +- ICDM +link: https://arxiv.org/abs/2605.00513 +author: Sitao Luan +categories: Publications +excerpt: We introduce ControBench, a benchmark for controversial discourse analysis that combines heterogeneous social interaction graphs with rich textual semantics, built from Reddit discussions on Trump, abortion and religion (7,370 users, 1,783 posts, 26,525 interactions). Its low or negative homophily reflects the cross-cutting structure of real debate, making it a challenging testbed for graph neural networks and language models. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Understanding how people argue across ideological divides online is important for studying political polarization, misinformation, and content moderation. Existing datasets capture only part of this problem: some preserve text but ignore interaction structure, some model structure without rich semantics, and others represent conversations without stable user-level ideological identity. We introduce ControBench, a benchmark for controversial discourse analysis that combines heterogeneous social interaction graphs with rich textual semantics. Built from Reddit discussions on three topics, Trump, abortion, and religion, ControBench contains 7,370 users, 1,783 posts, and 26,525 interactions. The graph contains user and post nodes connected by semantically enriched edges; in particular, user-comment-user edges encode both a reply and the parent comment that it responds to, preserving local argumentative context. User labels are derived from self-declared Reddit flairs, providing a scalable proxy for ideological identity without manual annotation. The resulting datasets exhibit low or negative adjusted homophily (Trump: -0.77, Abortion: 0.06, Religion: 0.04), reflecting the cross-cutting structure of real-world debate. We evaluate graph neural networks, pretrained language models, and large language models on ControBench and observe distinct performance patterns across topics and model families, especially when ideological boundaries are ambiguous. These results position ControBench as a challenging and realistic benchmark for controversial discourse analysis. diff --git a/_posts/papers/2026-11-01-emnlp2026-simulate-users.md b/_posts/papers/2026-11-01-emnlp2026-simulate-users.md new file mode 100644 index 000000000..7b76122db --- /dev/null +++ b/_posts/papers/2026-11-01-emnlp2026-simulate-users.md @@ -0,0 +1,16 @@ +--- +title: "Can LLMs Learn to Simulate Users from Behavioral History? An Empirical Study" +venue: The 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP) +names: Sneheel Sarangi, Aurélien Bück-Kaeffer, Maximilian Puelma Touzel, Zachary Yang, Jean-François Godbout, Reihaneh Rabbany +tags: +- EMNLP +author: Sneheel Sarangi +categories: Publications + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} diff --git a/_posts/teaching/2019-09-01-comp596-F19.md b/_posts/teaching/2019-09-01-comp599-F19.md similarity index 94% rename from _posts/teaching/2019-09-01-comp596-F19.md rename to _posts/teaching/2019-09-01-comp599-F19.md index 1d08a4a40..0d99f5ad4 100644 --- a/_posts/teaching/2019-09-01-comp596-F19.md +++ b/_posts/teaching/2019-09-01-comp599-F19.md @@ -1,7 +1,7 @@ --- title: Network Science term: Fall 2019 -course_code: COMP 596 +course_code: COMP 599 author: Reihaneh Rabbany names: Reihaneh Rabbany categories: Teaching @@ -16,7 +16,7 @@ tags: - Network Science --- -* **Course codes:** COMP 596 (Fall 2019) +* **Course codes:** COMP 599 (Fall 2019) * **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) * **Classroom:** ENGMC 103 * **Time:** Tuesdays and Thursdays, 10:00 am – 11:30 pm diff --git a/_posts/teaching/2020-01-01-comp551-W20.md b/_posts/teaching/2020-01-01-comp551-W20.md index a8f8cf24b..1ce37063b 100644 --- a/_posts/teaching/2020-01-01-comp551-W20.md +++ b/_posts/teaching/2020-01-01-comp551-W20.md @@ -3,7 +3,7 @@ title: Applied Machine Learning term: Winter 2020 course_code: COMP 551 author: Reihaneh Rabbany -names: Reihaneh Rabbany +names: Reihaneh Rabbany w/ Siamak Ravanbakhsh categories: Teaching share: false last_modified_at: 2024-10-11T00:00:00 @@ -17,7 +17,7 @@ tags: --- * **Course codes:** COMP 551 (Winter 2020) -* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) +* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/), Siamak Ravanbakhsh * **Classroom:** Strathcona Anatomy & Dentistry M-1 * **Time:** Tuesdays and Thursdays, 11:30 am - 12:55 am * **Course Website:** [here](http://www.reirab.com/Teaching/AML20/index.html) diff --git a/_posts/teaching/2020-09-01-comp596-F20.md b/_posts/teaching/2020-09-01-comp599-F20.md similarity index 95% rename from _posts/teaching/2020-09-01-comp596-F20.md rename to _posts/teaching/2020-09-01-comp599-F20.md index f7d597a05..f3acf9fbb 100644 --- a/_posts/teaching/2020-09-01-comp596-F20.md +++ b/_posts/teaching/2020-09-01-comp599-F20.md @@ -1,7 +1,7 @@ --- title: Network Science term: Fall 2020 -course_code: COMP 596 +course_code: COMP 599 author: Reihaneh Rabbany names: Reihaneh Rabbany categories: Teaching @@ -16,7 +16,7 @@ tags: - Network Science --- -* **Course codes:** COMP 596 (Fall 2020) +* **Course codes:** COMP 599 (Fall 2020) * **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) * **Location:** Remote through MyCourses * **Time:** Monday & Wednesday, 10:00 am – 11:30 pm diff --git a/_posts/teaching/2024-09-01-comp551-F24.md b/_posts/teaching/2024-09-01-comp551-F24.md index 695788596..d22bd04ab 100644 --- a/_posts/teaching/2024-09-01-comp551-F24.md +++ b/_posts/teaching/2024-09-01-comp551-F24.md @@ -1,9 +1,9 @@ --- title: Applied Machine Learning -term: Winter 2025 +term: Fall 2024 course_code: COMP 551 author: Reihaneh Rabbany -names: Reihaneh Rabbany +names: Reihaneh Rabbany w/ Isabeau Prémont-Schwarz categories: Teaching share: false last_modified_at: 2024-10-11T00:00:00 @@ -12,11 +12,11 @@ toc: true toc_sticky: true toc_label: "Table of Contents" tags: - - Winter 2025 + - Fall 2024 - Applied Machine Learning --- -* **Course codes:** COMP 551 (Winter 2025) +* **Course codes:** COMP 551 (Fall 2024) * **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/), [Isabeau Prémont-Schwarz](https://www.cs.mcgill.ca/~isabeau/) * **Location:** McConnell 204 * **Time:** Mondays and Wednesdays, 14:35-15:55 diff --git a/_posts/teaching/2025-01-01-comp511-W25.md b/_posts/teaching/2025-01-01-comp511-W25.md new file mode 100644 index 000000000..0ddcfc25b --- /dev/null +++ b/_posts/teaching/2025-01-01-comp511-W25.md @@ -0,0 +1,24 @@ +--- +title: Network Science +term: Winter 2025 +course_code: COMP 511 +author: Reihaneh Rabbany +names: Reihaneh Rabbany +categories: Teaching +share: false +last_modified_at: 2026-10-12T00:00:00 +layout: single +toc: true +toc_sticky: true +toc_label: "Table of Contents" +tags: + - Winter 2025 + - Network Science +--- + +* **Course codes:** COMP 511 (Winter 2025) +* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) + +# Overview + +This is on Network Science, Graph Mining and Graph Learning. Networks model relationships in complex systems, from hyperlinks between webpages, and co-authorships between research scholars to biological interactions between proteins and genes, and synaptic links between neurons. Network Science is an interdisciplinary research area involving researchers from Physics, Computer Science, Sociology, Math and Statistics, with applications in a wide range of domains including Biology, Medicine, Political Science, Marketing, Ecology, Criminology, etc. In this course, we will cover the basic concepts and techniques used in Network Science, review the state of the art techniques, and discuss the most recent developments. diff --git a/_posts/teaching/2025-01-01-comp551-W25.md b/_posts/teaching/2025-09-01-comp551-F25.md similarity index 63% rename from _posts/teaching/2025-01-01-comp551-W25.md rename to _posts/teaching/2025-09-01-comp551-F25.md index 93214a864..6847d6d29 100644 --- a/_posts/teaching/2025-01-01-comp551-W25.md +++ b/_posts/teaching/2025-09-01-comp551-F25.md @@ -1,26 +1,23 @@ --- -title: Applied Machine Learning -term: Winter 2025 +title: Applied Machine Learning +term: Fall 2025 course_code: COMP 551 author: Reihaneh Rabbany -names: Reihaneh Rabbany +names: Reihaneh Rabbany w/ Sékou-Oumar Kaba categories: Teaching share: false -last_modified_at: 2024-10-11T00:00:00 +last_modified_at: 2026-10-12T00:00:00 layout: single toc: true toc_sticky: true toc_label: "Table of Contents" tags: - - Winter 2025 + - Fall 2025 - Applied Machine Learning --- -* **Course codes:** COMP 551 (Winter 2025) -* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) -* **Location:** Stewart Biology Building S1/4 (lectures will be recorded) -* **Time:** Tuesdays and Thursdays, 1:05 pm - 2:25 pm -* **Course Website:** [here](http://www.reirab.com/Teaching/AML23/index.html) +* **Course codes:** COMP 551 (Fall 2025) +* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/), Sékou-Oumar Kaba # Overview diff --git a/_posts/teaching/2024-09-01-comp599-F24.md b/_posts/teaching/2026-01-01-comp511-W26.md similarity index 81% rename from _posts/teaching/2024-09-01-comp599-F24.md rename to _posts/teaching/2026-01-01-comp511-W26.md index 9c0a7a2ad..62269eadd 100644 --- a/_posts/teaching/2024-09-01-comp599-F24.md +++ b/_posts/teaching/2026-01-01-comp511-W26.md @@ -1,27 +1,25 @@ --- title: Network Science -term: Fall 2024 -course_code: COMP 599 +term: Winter 2026 +course_code: COMP 511 author: Reihaneh Rabbany names: Reihaneh Rabbany categories: Teaching share: false -last_modified_at: 2024-10-11T00:00:00 +last_modified_at: 2026-10-12T00:00:00 layout: single toc: true toc_sticky: true toc_label: "Table of Contents" tags: - - Fall 2024 + - Winter 2026 - Network Science --- -* **Course codes:** COMP 599 (Fall 2024) +* **Course codes:** COMP 511 (Winter 2026) * **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) # Overview This is on Network Science, Graph Mining and Graph Learning. Networks model relationships in complex systems, from hyperlinks between webpages, and co-authorships between research scholars to biological interactions between proteins and genes, and synaptic links between neurons. Network Science is an interdisciplinary research area involving researchers from Physics, Computer Science, Sociology, Math and Statistics, with applications in a wide range of domains including Biology, Medicine, Political Science, Marketing, Ecology, Criminology, etc. In this course, we will cover the basic concepts and techniques used in Network Science, review the state of the art techniques, and discuss the most recent developments. - -**Please note that about half of the class is paper discussions and student presentations.** \ No newline at end of file diff --git a/_research_directions/online-crime.md b/_research_directions/online-crime.md index c1b445f3b..8a7971aa3 100644 --- a/_research_directions/online-crime.md +++ b/_research_directions/online-crime.md @@ -18,6 +18,8 @@ logos: name: NSERC - image: /assets/images/logo/SAMSUNG.png name: SAMSUNG + - image: /assets/images/logo/IVADO.png + name: IVADO --- @@ -33,6 +35,6 @@ Sex trafficking impacts 4.8 million people globally and is a $99 billion USD ind # Funding -We acknowledge funding from the Canadian Institute for Advanced Research (CIFAR AI Chair Program), Mila - Quebec AI Institute, Natural Sciences and Engineering Research Council of Canada (NSERC), and Samsung. +We acknowledge funding from the Canadian Institute for Advanced Research (CIFAR AI Chair Program), Mila - Quebec AI Institute, Natural Sciences and Engineering Research Council of Canada (NSERC), Samsung, and IVADO (Institute for Data Valorization). {% include logo-grid.html id="logos" %} diff --git a/_research_directions/poli-sci.md b/_research_directions/poli-sci.md index 65aa26c2a..13ac954f4 100644 --- a/_research_directions/poli-sci.md +++ b/_research_directions/poli-sci.md @@ -24,6 +24,8 @@ logos: name: UKRI - image: /assets/images/logo/IVADO.png name: IVADO + - image: /assets/images/logo/AISI.png + name: AISI --- The current digital age fundamentally changes how we as individuals and a society collect and distribute information. With these changes come powerful new approaches to influence the public agenda. However, the quality of shared information varies widely, and can have wide-reaching consequences. As the sophistication and accessibility of AI tools continue to expand, the authenticity of digital content has been increasingly called into question. Our team stands at the forefront of the challenge to safeguard digital space, as we assess the nature of content and how it proliferates online. For more information check out our [a4.institute](https://www.ai4.institute/). @@ -38,6 +40,6 @@ The current digital age fundamentally changes how we as individuals and a societ # Funding -We acknowledge funding from the Canadian Institute for Advanced Research (CIFAR AI Chair Program), the Fonds de recherche du Québec – Société et culture (FRQSC), the Social Sciences and Humanities Research Council (SSHRC), Mila - Quebec AI Institute,the Government of Canada, UK Research and Innovation (UKRI), and IVADO (Institute for Data Valorization). +We acknowledge funding from the Canadian Institute for Advanced Research (CIFAR AI Chair Program), the Fonds de recherche du Québec – Société et culture (FRQSC), the Social Sciences and Humanities Research Council (SSHRC), Mila - Quebec AI Institute,the Government of Canada, UK Research and Innovation (UKRI), IVADO (Institute for Data Valorization), and the UK AI Security Institute (AISI). {% include logo-grid.html id="logos" %} \ No newline at end of file diff --git a/_sass/custom/display-highlighted-publications.scss b/_sass/custom/display-highlighted-publications.scss index 42486d2f0..89c5b82a5 100644 --- a/_sass/custom/display-highlighted-publications.scss +++ b/_sass/custom/display-highlighted-publications.scss @@ -1,30 +1,22 @@ +/* Selected publications on research direction pages */ + article.highlighted-publication-card { - display: grid; - grid-template-areas: - "title title" - "excerpt excerpt" - "venue venue" - "links links"; - grid-template-columns: 2fr 1fr; - column-gap: 15px; - margin-top: 1.5em; - align-items: start; - align-content: start; - - @media (max-width: 600px) { - grid-template-areas: - "title" - "excerpt" - "venue" - "links"; - grid-template-columns: 1fr; - } + display: flex; + flex-direction: column; + gap: 0.5rem; + padding: 1.25rem 1.5rem; + background: #fff; + border: 1px solid rgba(0, 0, 0, 0.07); + border-radius: 12px; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04); } .highlighted-publication-title { - grid-area: title; - margin: 0 0 5px 0; + margin: 0; + padding: 0; border-bottom: none !important; + font-size: 1.05rem; + line-height: 1.35; >a { text-decoration: none; @@ -32,9 +24,9 @@ article.highlighted-publication-card { } .highlighted-publication-excerpt { - grid-area: excerpt; - display: block; - font-style: italic; + font-size: 0.8rem; + line-height: 1.6; + color: #555; >p { margin: 0; @@ -42,12 +34,13 @@ article.highlighted-publication-card { } .highlighted-publication-venue { - grid-area: venue; - font-weight: bold; + font-size: 0.75rem; + font-weight: 600; + color: #333; } -.publication-links { - grid-area: links; +article.highlighted-publication-card .publication-links { + font-size: 0.75rem; } .publication-ext-link { @@ -57,5 +50,78 @@ article.highlighted-publication-card { } .highlighted-publications-spacing { - margin-bottom: 1rem; // Adjust this value to set the desired spacing -} \ No newline at end of file + height: 1rem; +} + +/* "Show N more" toggle under the first few selected publications (posts-highlighted-publications.html): + a floating pill centred on a hairline divider. */ +.highlighted-publications__toggle-row { + display: flex; + align-items: center; + gap: 1rem; + margin: 0.75rem 0 0; + + &::before, + &::after { + content: ""; + flex: 1; + height: 1px; + background: linear-gradient(to right, transparent, rgba(0, 0, 0, 0.1)); + } + + &::after { + background: linear-gradient(to left, transparent, rgba(0, 0, 0, 0.1)); + } +} + +.highlighted-publications__toggle { + display: inline-flex; + align-items: center; + gap: 0.6rem; + margin: 0; + padding: 0.4rem 0.45rem 0.4rem 1.1rem; + border: 1px solid rgba(0, 0, 0, 0.08); + border-radius: 999px; + background: #fff; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04), 0 4px 14px rgba(0, 0, 0, 0.06); + color: #333; + font-family: inherit; + font-size: 0.8rem; + font-weight: 600; + letter-spacing: 0.01em; + cursor: pointer; + transition: box-shadow 0.2s ease, transform 0.2s ease, color 0.2s ease; + + &:hover, + &:focus-visible { + color: $link-color; + transform: translateY(-1px); + box-shadow: 0 2px 4px rgba(0, 0, 0, 0.05), 0 8px 22px rgba(0, 0, 0, 0.09); + } + + &:focus-visible { + outline: 2px solid rgba($link-color, 0.35); + outline-offset: 2px; + } +} + +.highlighted-publications__toggle-icon { + display: inline-flex; + align-items: center; + justify-content: center; + width: 1.6rem; + height: 1.6rem; + border-radius: 50%; + background: rgba($link-color, 0.1); + color: $link-color; + font-size: 0.65rem; + transition: transform 0.25s ease, background 0.2s ease; + + .highlighted-publications__toggle:hover & { + background: rgba($link-color, 0.16); + } + + .highlighted-publications__toggle[aria-expanded="true"] & { + transform: rotate(180deg); + } +} diff --git a/_sass/custom/display-publications.scss b/_sass/custom/display-publications.scss index daff0e085..5d32e78c0 100644 --- a/_sass/custom/display-publications.scss +++ b/_sass/custom/display-publications.scss @@ -1,31 +1,38 @@ +/* Mirrors the selected-publication cards on research pages (display-highlighted-publications.scss), + with the author list in place of the summary. */ article.publication-card { - display: grid; - grid-template-areas: - "thumbnail title" - "thumbnail authors" - "thumbnail venue" - "thumbnail links"; - grid-template-columns: 300px 1fr; - gap: 10px; - margin: 0.75em 0; - padding: 1em; + display: flex; + flex-direction: column; + gap: 0.5rem; + margin: 0 0 1rem; + padding: 1.25rem 1.5rem; background: #fff; - border-radius: 4px; - box-shadow: 0 1px 3px rgba(0, 0, 0, 0.12); - transition: box-shadow 0.3s ease; + border: 1px solid rgba(0, 0, 0, 0.07); + border-radius: 12px; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04); - &:hover { - box-shadow: 0 3px 6px rgba(0, 0, 0, 0.16); - } - - @media (max-width: 600px) { + // Only reserve an image column for papers that set `thumbnail` in front matter. + &--with-thumbnail { + display: grid; grid-template-areas: - "title" - "authors" - "venue" - "links" - "thumbnail"; - grid-template-columns: 1fr; + "thumbnail title" + "thumbnail authors" + "thumbnail venue" + "thumbnail award" + "thumbnail links"; + grid-template-columns: 300px 1fr; + gap: 10px; + + @media (max-width: 600px) { + grid-template-areas: + "title" + "authors" + "venue" + "award" + "links" + "thumbnail"; + grid-template-columns: 1fr; + } } } @@ -40,22 +47,22 @@ article.publication-card { .publication-title { grid-area: title; margin: 0; + padding: 0; + border-bottom: none; + font-size: 1.05rem; + line-height: 1.35; >a { - color: $primary-color; text-decoration: none; - - &:hover { - color: darken($primary-color, 10%); - text-decoration: none; - } } } .publication-authors { grid-area: authors; - font-style: italic; - margin: 0; + margin: 0.5rem 0 0; + font-size: 0.8rem; + line-height: 1.6; + color: #555; white-space: normal; overflow: visible; word-wrap: break-word; @@ -63,8 +70,10 @@ article.publication-card { .publication-venue { grid-area: venue; - font-weight: bold; margin: 0; + font-size: 0.75rem; + font-weight: 600; + color: #333; } .publication-links { @@ -141,4 +150,47 @@ article.publication-card { margin-top: 2em; padding-bottom: 0.5em; border-bottom: 1px solid #f2f3f3; -} \ No newline at end of file +} +/* Award note after the venue (set `award:` and optionally `award_url:` in a paper's front matter) */ +.publication-award { + font-style: italic; + font-weight: normal; + + a { + color: inherit; + text-decoration: none; + + &:hover { + text-decoration: underline; + } + } +} + +/* Back link on individual paper pages (_layouts/publication.html); a script moves it above the title. */ +.back-to-publications { + display: inline-flex; + align-items: center; + gap: 0.4rem; + margin-bottom: 0.75rem; + font-size: 0.8rem; + font-weight: 600; + color: #555; + text-decoration: none !important; + + &:hover, + &:focus-visible { + color: $primary-color; + } +} + +/* Paper / Code / Page links: same plain style as the research-page cards */ +article.publication-card .publication-links { + font-size: 0.75rem; +} + +/* Publications page: comfortable reading width, like the research pages */ +.publications-page #main>.splash { + max-width: 960px; + margin: 0 auto; + padding: 0 1rem; +} diff --git a/_sass/custom/display-tech-transfer.scss b/_sass/custom/display-tech-transfer.scss index 10622c45d..7cda94ffe 100644 --- a/_sass/custom/display-tech-transfer.scss +++ b/_sass/custom/display-tech-transfer.scss @@ -1,32 +1,192 @@ +/* Tools & resources entries (Tools & Resources page and research direction pages) */ + .tech-transfer__wrapper { + display: flex; + flex-direction: column; + gap: 1rem; margin-bottom: 2em; .tech-transfer-item { - margin-bottom: 3em; - padding: 1em; + padding: 1.5rem; background: #fff; - border-radius: 4px; - transition: all 0.3s cubic-bezier(.25, .8, .25, 1); + border: 1px solid rgba(0, 0, 0, 0.07); + border-radius: 12px; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04); h2 { - margin-top: 0; + margin: 0 0 1rem; + padding: 0; + border-bottom: none; + font-size: 1.2rem; color: $primary-color; + + a { + text-decoration: none; + } + } + + p { + margin: 0 0 0.75rem; + font-size: 0.9rem; + line-height: 1.65; + + &:last-child { + margin-bottom: 0; + } } img { - max-width: 200px; + // Never upscale small logos; cap large images at 200px. + width: auto; + max-width: min(200px, 40%); float: left; - margin: 0 1em 1em 0; + margin: 0.2rem 1.5rem 0.5rem 0; + border-radius: 6px; } .clearfix { - clear: both; - &::after { content: ""; clear: both; display: table; } } + + @media (max-width: 600px) { + img { + float: none; + display: block; + width: 100%; + max-width: 100%; + margin: 0 0 1rem; + } + } + } +} + +/* Link buttons and pip install chip under each entry */ + +.tool-links { + clear: both; + display: flex; + flex-wrap: wrap; + gap: 0.5rem; + margin-top: 1.25rem; +} + +.tool-link, +.tool-pip { + display: inline-flex; + align-items: center; + gap: 0.5rem; + padding: 0.45rem 0.9rem; + border: 1px solid rgba(155, 58, 43, 0.25); + border-radius: 999px; + background: #fff; + color: #9b3a2b; + font-size: 0.75rem; + font-weight: 600; + line-height: 1.2; + text-decoration: none !important; + transition: background-color 0.15s ease, border-color 0.15s ease, color 0.15s ease; + + &:hover, + &:focus-visible { + background: #f6efed; + border-color: rgba(155, 58, 43, 0.45); + color: #9b3a2b; + } +} + +.tool-link--primary { + background: #9b3a2b; + border-color: #9b3a2b; + color: #fff; + + &:hover, + &:focus-visible { + background: #7f2f23; + border-color: #7f2f23; + color: #fff; + } +} + +.tool-pip { + border-color: #e4e4e4; + background: #f7f7f8; + color: #333; + cursor: pointer; + font-family: inherit; + + code { + padding: 0; + background: none; + font-size: 0.75rem; + color: inherit; + } + + .tool-pip__icon { + color: #999; + } + + &:hover, + &:focus-visible { + background: #efeff1; + border-color: #d6d6d6; + color: #333; + } + + &.is-copied { + border-color: #6aa36f; + + .tool-pip__icon::before { + content: "\f00c"; // check mark + font-weight: 900; + color: #3d8b44; + } + } +} + +/* Sort dropdown above the Tools & Resources list */ + +.tool-sort { + display: flex; + justify-content: flex-end; + margin: 0 0 0.5rem; + + &__control { + position: relative; + display: inline-flex; + align-items: center; + color: #555; + font-size: 0.8rem; + + .fa-chevron-down { + position: absolute; + right: 0.6rem; + font-size: 0.65rem; + pointer-events: none; + } + } + + &__select { + appearance: none; + -webkit-appearance: none; + margin: 0; + padding: 0.35rem 1.8rem 0.35rem 0.6rem; + border: none; + border-radius: 6px; + background: transparent; + color: inherit; + font: inherit; + font-weight: 600; + cursor: pointer; + box-shadow: none; + + &:hover, + &:focus-visible { + background: #f3f3f3; + outline: none; + } } -} \ No newline at end of file +} diff --git a/_sass/custom/funders-carousel.scss b/_sass/custom/funders-carousel.scss new file mode 100644 index 000000000..524f2f476 --- /dev/null +++ b/_sass/custom/funders-carousel.scss @@ -0,0 +1,94 @@ +/* + Funders carousel (homepage) + ========================================================================== */ + +@keyframes fundersScroll { + from { + transform: translateX(0); + } + + to { + // The track holds the logo list twice, so shifting by half loops seamlessly. + transform: translateX(-50%); + } +} + +.funders-carousel { + margin: 6rem 0 3rem; + text-align: center; + + &__title { + margin-bottom: 1.5rem; + padding-bottom: 0; + border-bottom: none; + font-size: $type-size-4; + color: $text-color; + } + + &__viewport { + max-width: 760px; + margin: 0 auto; + overflow: hidden; + mask-image: linear-gradient(to right, transparent, #000 10%, #000 90%, transparent); + -webkit-mask-image: linear-gradient(to right, transparent, #000 10%, #000 90%, transparent); + } + + &__track { + display: flex; + width: max-content; + margin: 0; + padding: 0; + list-style: none; + animation: fundersScroll 40s linear infinite; + } + + &__item { + flex: 0 0 auto; + display: flex; + align-items: center; + justify-content: center; + height: 70px; + margin: 0 2.5rem; + + img { + max-height: 100%; + max-width: 160px; + width: auto; + object-fit: contain; + // filter/opacity are set per frame by the script in funders-carousel.html + will-change: filter, opacity; + } + } + + @media (max-width: 768px) { + &__item { + height: 50px; + margin: 0 1.5rem; + + img { + max-width: 110px; + } + } + } + + // No auto-scroll for reduced motion: show one copy and let users swipe/scroll. + @media (prefers-reduced-motion: reduce) { + &__viewport { + overflow-x: auto; + } + + &__track { + animation: none; + } + + &__item[aria-hidden="true"] { + display: none; + } + } +} + +// Override the theme's underline on content h2s +.page__content .funders-carousel__title { + padding-bottom: 0; + border-bottom: none; +} diff --git a/_sass/custom/header-footer.scss b/_sass/custom/header-footer.scss index 0909ac243..b1a896570 100644 --- a/_sass/custom/header-footer.scss +++ b/_sass/custom/header-footer.scss @@ -1,19 +1,63 @@ +$navbar-bg-color: #faf9f6; +$navbar-text-color: #9b3a2b; + a.site-title { + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; + @media (min-width: 601px) { - font-size: xx-large; + font-size: 1.35rem; } @media (max-width: 600px) { - font-size: large; + font-size: 0.95rem; } - color: $primary-color; + color: $navbar-text-color; + line-height: 1.1; &:hover { - color: mix($background-color, $primary-color, 25%); + color: mix(#000, $navbar-text-color, 20%); } } +.masthead, +.greedy-nav { + background-color: $navbar-bg-color; +} + +.masthead { + padding-top: 0.15rem; + padding-bottom: 0.15rem; +} + +.masthead .masthead__inner-wrap { + max-width: 100%; + width: 100%; + padding-top: 0.15rem; + padding-bottom: 0.15rem; + padding-left: 1rem; + padding-right: 1rem; +} + +.greedy-nav { + min-height: 2.6rem; + line-height: 2.6rem; +} + +.greedy-nav .visible-links a { + padding-top: 0; + padding-bottom: 0; + line-height: 2.6rem; +} + +.greedy-nav a { + color: $navbar-text-color; +} + +.greedy-nav a:hover { + color: mix(#000, $navbar-text-color, 20%); +} + .theme-toggle { @media (min-width: 601px) { margin: 0px; @@ -21,22 +65,72 @@ a.site-title { } +/* Dropdown menus + ========================================================================== */ + .submenu { list-style: none; } -.masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li:hover>.submenu { +.masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li:hover>.submenu, +.masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li:focus-within>.submenu { display: block; } .masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li>.submenu { position: absolute; - background: $background-color; - min-width: 200px; - z-index: 1000; display: none; + z-index: 1000; + min-width: 250px; + margin: 0 0 0 0.25rem; + padding: 0.5rem; + background: #fff; + border: 1px solid rgba(0, 0, 0, 0.06); + border-radius: 10px; + box-shadow: 0 10px 30px rgba(0, 0, 0, 0.08), 0 2px 6px rgba(0, 0, 0, 0.04); + line-height: 1.4; + + >li>a { + margin: 0; + padding: 0.5rem 0.75rem; + border-radius: 6px; + font-size: 0.9rem; + line-height: 1.4; + color: #2b2b2b; + white-space: nowrap; + transition: background-color 0.15s ease, color 0.15s ease; + + // No navbar underline animation inside the dropdown + &::before { + display: none; + } + + &:hover, + &:focus-visible { + background: #f6efed; + color: $navbar-text-color; + } + } +} + +// Section label and separator (`heading:` / `divider:` entries in _data/navigation.yml) +.submenu__heading { + padding: 0.5rem 0.75rem 0.35rem; + font-size: 0.68rem; + font-weight: 600; + letter-spacing: 0.12em; + text-transform: uppercase; + line-height: 1.4; + color: $gray; + cursor: default; +} + +.submenu__divider { + height: 1px; + margin: 0.4rem 0.75rem; + background: #ececec; } -.masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li>.submenu>li>a { - padding: 10px 20px; +.greedy-nav .hidden-links .submenu__divider { + border-bottom: none; } \ No newline at end of file diff --git a/_sass/custom/particles.scss b/_sass/custom/particles.scss index b4c606dcb..c2cc54fca 100644 --- a/_sass/custom/particles.scss +++ b/_sass/custom/particles.scss @@ -12,9 +12,11 @@ .page__hero--overlay { position: relative; - height: 300px; + // Grow with the content instead of a fixed height, so titles never crowd the navbar. + min-height: 300px; + background-color: #fff; overflow: hidden; - padding: 0 1em; + padding: 4rem 1em 3rem; display: flex; align-items: center; justify-content: center; @@ -53,7 +55,7 @@ margin: 0 auto; @media (min-width: 80em) { - padding: 3em; + padding: 0 3em; } @media (max-width: 768px) { @@ -65,12 +67,13 @@ .page_overlay_image { height: 100%; - min-height: 300px; + min-height: 220px; } .page__title-container { animation: fadeInUp 0.6s ease forwards; grid-column: 2; + position: relative; .title-line { width: 100px; @@ -91,6 +94,7 @@ .page__title { font-size: $type-size-1; + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #000; margin: 0; line-height: 1.2; @@ -102,4 +106,70 @@ color: #000; margin-top: 1em; } +} + +.page__hero--particles { + // Research direction pages: icon inline before the title, text below, + // aligned with the 960px content column (see research-page.scss). + .wrapper--research { + display: block; + max-width: 960px; + padding: 0 1rem; + } + + .page__title-logo { + display: inline-block; + height: 1em; + width: auto; + margin-right: 0.35em; + vertical-align: -0.12em; + } + + .wrapper, + .page_overlay_image, + .page__title-container { + pointer-events: none; + } + + .page__title-container { + &::before { + content: ""; + position: absolute; + top: 50%; + left: 50%; + width: 118%; + height: 145%; + transform: translate(-50%, -50%); + background: radial-gradient(ellipse at center, + rgba(255, 255, 255, 1) 0%, + rgba(255, 255, 255, 0.97) 25%, + rgba(255, 255, 255, 0.8) 52%, + rgba(255, 255, 255, 0.35) 72%, + rgba(255, 255, 255, 0.1) 82%, + rgba(255, 255, 255, 0) 90%); + filter: blur(2px); + z-index: -1; + pointer-events: none; + } + } + + .page__title, + .page__lead, + .title-line { + position: relative; + z-index: 2; + } + + .page__title, + .page__lead { + color: #000 !important; + opacity: 1; + } +} + + + +h1.page__title, +.page__content>h1:first-of-type { + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; } \ No newline at end of file diff --git a/_sass/custom/people-card.scss b/_sass/custom/people-card.scss index 009a26d22..af20bd1ff 100644 --- a/_sass/custom/people-card.scss +++ b/_sass/custom/people-card.scss @@ -244,4 +244,9 @@ div.author-avatar { margin-right: .2em; font-size: x-large; color: $primary-color -} \ No newline at end of file +} +/* People page: extra room before the Past Members section */ +.page__content hr + #past-members, +#past-members { + margin-top: 4.5rem; +} diff --git a/_sass/custom/research-directions.scss b/_sass/custom/research-directions.scss index 1aaa71f75..6e9459ffb 100644 --- a/_sass/custom/research-directions.scss +++ b/_sass/custom/research-directions.scss @@ -2,6 +2,51 @@ Research Directions ========================================================================== */ +@keyframes logoFloat { + 0%, + 100% { + transform: translateY(0); + } + + 50% { + transform: translateY(-5px); + } +} + +.research-directions-section { + position: relative; +} + +.research-directions-section--with-particles { + margin-top: 2.5rem; + width: 100vw; + margin-left: calc(50% - 50vw); + margin-right: calc(50% - 50vw); + + #particles-js { + position: absolute; + top: 0; + left: 0; + width: 100%; + height: 100%; + z-index: 0; + background-color: #faf9f6; + } + + .research-directions__wrapper { + position: relative; + z-index: 1; + max-width: 1200px; + margin-left: auto; + margin-right: auto; + padding-top: 1.5em; + padding-left: 1em; + padding-right: 1em; + border-radius: 10px; + pointer-events: none; + } +} + .research-directions__wrapper { @include clearfix(); padding-bottom: 1em; @@ -42,6 +87,151 @@ } } +.research-directions-item--home { + box-shadow: none; + border-radius: 0; + min-height: 0; + padding-bottom: 1.1rem; + pointer-events: none; + + &:hover { + transform: none; + } + + a { + position: relative; + display: flex; + flex-direction: column; + align-items: center; + justify-content: flex-start; + text-align: center; + height: auto; + overflow: visible; + background: transparent; + padding: 1.1rem 0.5rem 0.35rem; + isolation: isolate; + pointer-events: auto; + } + + .direction-logo-wrap { + position: relative; + z-index: 2; + display: inline-flex; + align-items: center; + justify-content: center; + margin-bottom: 0.9rem; + animation: logoFloat 4.4s ease-in-out infinite; + + &::before { + content: ""; + position: absolute; + top: 50%; + left: 50%; + width: 150px; + height: 150px; + transform: translate(-50%, -50%); + background: radial-gradient(ellipse at center, + rgba(255, 255, 255, 1) 0%, + rgba(255, 255, 255, 0.98) 20%, + rgba(255, 255, 255, 0.85) 42%, + rgba(255, 255, 255, 0.45) 62%, + rgba(255, 255, 255, 0.14) 76%, + rgba(255, 255, 255, 0) 86%); + filter: blur(2px); + z-index: 0; + pointer-events: none; + } + } + + @media (prefers-reduced-motion: reduce) { + .direction-logo-wrap { + animation: none; + } + } + + .direction-logo { + position: relative; + z-index: 3; + width: 95px; + height: 95px; + object-fit: contain; + transition: transform 220ms ease; + box-shadow: none !important; + filter: none !important; + } + + a:hover .direction-logo, + a:focus .direction-logo, + a:focus-visible .direction-logo { + transform: scale(1.08); + } + + a:hover .direction-logo, + a:focus .direction-logo, + a:active .direction-logo { + box-shadow: none !important; + filter: none !important; + } + + .direction-content { + position: relative; + z-index: 1; + max-width: 320px; + margin-left: auto; + margin-right: auto; + padding: 0.35rem 0.45rem 0.5rem; + transition: transform 220ms ease; + transform-origin: center top; + + &::before { + content: ""; + position: absolute; + top: 50%; + left: 50%; + width: 140%; + height: 150%; + transform: translate(-50%, -50%); + background: radial-gradient(ellipse at center, + rgba(255, 255, 255, 1) 0%, + rgba(255, 255, 255, 0.98) 26%, + rgba(255, 255, 255, 0.82) 50%, + rgba(255, 255, 255, 0.34) 70%, + rgba(255, 255, 255, 0.1) 78%, + rgba(255, 255, 255, 0) 86%); + filter: blur(2px); + z-index: -1; + pointer-events: none; + } + } + + a:hover .direction-content, + a:focus .direction-content, + a:focus-visible .direction-content { + transform: scale(1.04); + } + + .direction-title { + justify-content: center; + border-bottom: 0; + padding-bottom: 0.2rem; + + h2 { + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; + font-size: 1.35rem; + line-height: 1.08; + } + } + + .direction-excerpt { + margin-top: 0; + font-size: 0.98rem; + + p { + line-height: 1.3; + } + } +} + .direction-image { position: relative; min-height: 250px; diff --git a/_sass/custom/research-links.scss b/_sass/custom/research-links.scss new file mode 100644 index 000000000..90992c52a --- /dev/null +++ b/_sass/custom/research-links.scss @@ -0,0 +1,88 @@ +/* + Research links (homepage cards below the research directions) + ========================================================================== */ + +html { + scroll-behavior: smooth; +} + +#research { + scroll-margin-top: 1rem; +} + +@media (prefers-reduced-motion: reduce) { + html { + scroll-behavior: auto; + } +} + +.research-links { + display: grid; + grid-template-columns: repeat(3, 1fr); + gap: 1.25rem; + max-width: 1100px; + margin: 6rem auto; + + @media (max-width: 900px) { + grid-template-columns: 1fr; + max-width: 560px; + } + + &__card, + &__card:hover, + &__card:focus-visible, + &__card * { + text-decoration: none !important; + } + + &__card { + display: flex; + align-items: center; + gap: 1rem; + padding: 1.25rem 1.5rem; + background: #fff; + border: 1px solid rgba(0, 0, 0, 0.07); + border-radius: 14px; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04); + color: inherit; + transition: transform 0.2s ease, box-shadow 0.2s ease, border-color 0.2s ease; + + &:hover, + &:focus-visible { + transform: translateY(-3px); + border-color: rgba(155, 58, 43, 0.25); + box-shadow: 0 12px 28px rgba(0, 0, 0, 0.08); + color: inherit; + } + } + + &__icon { + flex: 0 0 auto; + display: flex; + align-items: center; + justify-content: center; + width: 2.5rem; + height: 2.5rem; + border-radius: 10px; + background: #f6efed; + color: #9b3a2b; + font-size: 1rem; + } + + &__title { + font-weight: 700; + font-size: 1rem; + line-height: 1.3; + color: #1f1f1f; + } + + @media (prefers-reduced-motion: reduce) { + &__card { + transition: none; + } + + &__card:hover { + transform: none; + } + } +} diff --git a/_sass/custom/research-page.scss b/_sass/custom/research-page.scss new file mode 100644 index 000000000..a22752860 --- /dev/null +++ b/_sass/custom/research-page.scss @@ -0,0 +1,31 @@ +/* + Research direction pages (body.research-page, set in _config.yml defaults) + ========================================================================== */ + +.research-page { + .splash>.page__content { + max-width: 960px; + margin: 0 auto; + padding: 0 1rem 2rem; + + >p { + line-height: 1.75; + } + + >h1 { + margin: 3.5rem 0 1.25rem; + padding-bottom: 0.6rem; + border-bottom: 1px solid #ececec; + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; + font-size: 1.5rem; + } + + >h1:first-of-type { + margin-top: 2.5rem; + } + } + + .logo-grid { + padding: 1.25rem 0; + } +} diff --git a/_sass/minimal-mistakes.scss b/_sass/minimal-mistakes.scss index 969675c81..c92eb03f5 100644 --- a/_sass/minimal-mistakes.scss +++ b/_sass/minimal-mistakes.scss @@ -52,7 +52,10 @@ @import "custom/research-directions.scss"; @import "custom/sub_research_directions.scss"; @import "custom/team-grid.scss"; -@import "custom/logo-grid.scss"; +@import "custom/logo-grid.scss"; +@import "custom/funders-carousel.scss"; +@import "custom/research-links.scss"; +@import "custom/research-page.scss"; @import "custom/display-highlighted-publications.scss"; @import "custom/breadcrumbs.scss"; @import "custom/particles.scss"; \ No newline at end of file diff --git a/_tech_transfer/credibench.md b/_tech_transfer/credibench.md new file mode 100644 index 000000000..63710ebb8 --- /dev/null +++ b/_tech_transfer/credibench.md @@ -0,0 +1,23 @@ +--- +category: information-integrity +paper_date: 2026-08-09 # date of the most recent paper; entries are listed newest first +pip: credigraph +links: + - label: Domain query API docs + url: https://credinet.readthedocs.io/en/latest/ + type: docs + - label: Dataset + url: https://huggingface.co/datasets/ComplexDataLab/CrediBench + type: dataset + - label: Paper (KDD 2026) + url: https://dl.acm.org/doi/10.1145/3770855.3817476 + type: paper +--- + +## [CrediBench](https://dl.acm.org/doi/10.1145/3770855.3817476) + +
# Join Us diff --git a/_pages/home.md b/_pages/home.md index 577191d52..a8d55417c 100644 --- a/_pages/home.md +++ b/_pages/home.md @@ -7,6 +7,27 @@ header: overlay_css: "background-size: contain; background-repeat: no-repeat;" excerpt: "A research group focusing on developing techniques for analyzing complex data from online societies, with applications to enhance the health and safety of online spaces." +# Homepage funders strip: logos are collected from the research directions' `logos`; +# `exclude` drops some (by image path) and `extra` adds funders not listed on a research page. +funders: + exclude: + - /assets/images/logo/GOOGLE.png + - /assets/images/logo/SAMSUNG.png + extra: + - image: /assets/images/logo/FLI.png + name: Future of Life Institute + +research_links: + - title: "Tools & Resources" + url: /tools-and-resources/ + icon: "fas fa-toolbox" + - title: "Publications" + url: /publications/ + icon: "fas fa-book-open" + - title: "STAMINA Reading Group" + url: /stamina/ + icon: "fas fa-users" + row_research: - image_path: /assets/images/home/group-photo-2.webp url: /publications @@ -20,4 +41,8 @@ row_research: {% include research-directions.html category="research-directions"%} +{% include research-links.html %} + {% include feature_row id="row_research" type="left" %} + +{% include funders-carousel.html %} diff --git a/_pages/people.md b/_pages/people.md index d1485671f..e4d9e777f 100644 --- a/_pages/people.md +++ b/_pages/people.md @@ -7,7 +7,7 @@ header: overlay_css: "background-size: contain; background-repeat: no-repeat;" --- -## Faculty & Staff & Postdoc +## Faculty, Staff & Postdoc {% include card-authors-with-role.html authors=site.data.authors role="Faculty" render_current_role=true %} {% include card-authors-with-role.html authors=site.data.authors role="Staff" render_current_role=true %} @@ -15,49 +15,31 @@ header: ## Students -### PhD +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="PhD" heading="PhD" heading_id="phd" render_current_role=true %} -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="PhD" render_current_role=true %} +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Master" heading="Masters" heading_id="masters" render_current_role=true %} -### Masters +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Undergraduate" heading="Undergraduate" heading_id="undergraduate" render_current_role=true %} -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Master" render_current_role=true %} - -### Undergraduate - -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Undergraduate" render_current_role=true %} - -### Intern - -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Intern" render_current_role=true %} +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Intern" heading="Intern" heading_id="intern" render_current_role=true %} --- # Past Members -## Faculty & Staff & Postdoc +## Faculty, Staff & Postdocs -{% include card-authors-with-role.html authors=site.data.authors role="Faculty" render_current_role=false %} -{% include card-authors-with-role.html authors=site.data.authors role="Staff" render_current_role=false %} -{% include card-authors-with-role.html authors=site.data.authors role="Postdoc" render_current_role=false %} +{% include card-authors-with-role.html authors=site.data.authors roles="Faculty,Staff,Postdoc" render_current_role=false %} ## Alumni -### PhD - -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="PhD" render_current_role=false %} - -### Masters - -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Master" render_current_role=false %} - -### Undergraduate +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="PhD" heading="PhD" heading_id="phd-past" render_current_role=false %} -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Undergraduate" render_current_role=false %} +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Master" heading="Masters" heading_id="masters-past" render_current_role=false %} -### Interns +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Undergraduate" heading="Undergraduate" heading_id="undergraduate-past" render_current_role=false %} -{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Intern" render_current_role=false %} +{% include card-authors-with-role.html authors=site.data.authors role="Student" title="Intern" heading="Interns" heading_id="interns-past" render_current_role=false %}
diff --git a/_pages/publications.md b/_pages/publications.md index c8af1c166..b1c68065f 100644 --- a/_pages/publications.md +++ b/_pages/publications.md @@ -8,6 +8,7 @@ excerpt: "" classes: - no-sidebar - wide + - publications-page permalink: /publications/ taxonomy: Publications show_view_by_venue: true diff --git a/_pages/tools-and-resources.md b/_pages/tools-and-resources.md index 4ddd8dac8..b469f0020 100644 --- a/_pages/tools-and-resources.md +++ b/_pages/tools-and-resources.md @@ -10,4 +10,4 @@ excerpt: "Showcasing our tools and resources that bridge research innovations wi --- -{% include tech-transfer.html %} +{% include tech-transfer.html sortable=true %} diff --git a/_posts/news/2025-06-29-TGL_2025.md b/_posts/news/2025-06-29-TGL_2025.md index 923b87a26..eb85d1154 100644 --- a/_posts/news/2025-06-29-TGL_2025.md +++ b/_posts/news/2025-06-29-TGL_2025.md @@ -1,6 +1,6 @@ --- title: "TGL'25" -venue: Temporal Graph Learning Workshop (Third Edition) @ KDD 2025! +venue: Temporal Graph Learning Workshop (Third Edition) @ KDD 2025 names: Farimah Poursafaei, Shenyang Huang, Julia Gastinger, Daniele Zambon, Andrea Cini, Jacob Chmura, Reihaneh Rabbany, Michael Bronstein author: Farimah Poursafaei categories: News diff --git a/_posts/news/2025-07-01-Social_Sim_2025.md b/_posts/news/2025-07-01-Social_Sim_2025.md index 06823b2e5..86f1a3640 100644 --- a/_posts/news/2025-07-01-Social_Sim_2025.md +++ b/_posts/news/2025-07-01-Social_Sim_2025.md @@ -1,6 +1,6 @@ --- title: "Social Sim'25" -venue: Social Simulation with LLMs has been accepted at COLM 2025! +venue: Social Simulation with LLMs has been accepted at COLM 2025 names: Zachary Yang, Kellin Pelrine, Maximillian Puelma Touzel, Dan Zhao, Jean-Francois Godbout, Petter Törnberg, Reihaneh Rabbany, William Cunningham author: Zachary Yang categories: News diff --git a/_posts/news/2025-07-09-FIIR_2025.md b/_posts/news/2025-07-09-FIIR_2025.md index cf056201d..447afeace 100644 --- a/_posts/news/2025-07-09-FIIR_2025.md +++ b/_posts/news/2025-07-09-FIIR_2025.md @@ -1,6 +1,6 @@ --- title: "FIIR'25" -venue: Future of Information Integrity Research will be co-located with ICDM'25! +venue: Future of Information Integrity Research will be co-located with ICDM'25 names: Reihaneh Rabbany, Kai Shu, Jean-Francois Godbout, Taylor Lynn Curtis, Bijean Ghafouri, Kellin Pelrine, Maximilian Puelma Touzel author: Reihaneh Rabbany categories: News diff --git a/_posts/news/2026-10-09-Social_Sim_2026.md b/_posts/news/2026-10-09-Social_Sim_2026.md new file mode 100644 index 000000000..fdb632e0e --- /dev/null +++ b/_posts/news/2026-10-09-Social_Sim_2026.md @@ -0,0 +1,11 @@ +--- +title: "Social Sim'26" +venue: The 2nd Workshop on Social Simulation with LLMs at COLM 2026 +names: Zachary Yang, Maximilian Puelma Touzel, Xuhui Zhou, Yunze Xiao, Lynnette Hui Xian Ng, Jean-Francois Godbout, Reihaneh Rabbany, Maarten Sap +author: Zachary Yang +categories: News +thumbnail: "assets/images/news/social_sim.jfif" +--- + + +We’re pleased to announce Social Sim’26, the 2nd Workshop on Social Simulation with LLMs, taking place at COLM 2026 in San Francisco on October 9. This year’s theme is *Fidelity in Applications*. The workshop also hosts the *Many Worlds* shared task on evaluating LLM social simulations, with [code](https://github.com/sandbox-social/socsim26_sharedtask) and [data](https://huggingface.co/datasets/ComplexDataLab/socsim26-sharedtask) openly available. For more information, please refer to the [workshop website](https://sites.google.com/view/social-sims-with-llms/). diff --git a/_posts/papers/2024-11-07-2411.05060.md b/_posts/papers/2024-11-07-2411.05060.md deleted file mode 100644 index 58383d5cf..000000000 --- a/_posts/papers/2024-11-07-2411.05060.md +++ /dev/null @@ -1,28 +0,0 @@ ---- -title: A Guide to Misinformation Detection Datasets -venue: arXiv.org -names: Camille Thibault, Gabrielle Peloquin-Skulski, Jacob-Junqi Tian, Florence Laflamme, - Yuxiang Guan, Reihaneh Rabbany, J. Godbout, Kellin Pelrine -tags: -- arXiv.org -link: https://arxiv.org/abs/2411.05060 -webpage: https://misinfo-datasets.complexdatalab.com/ -code: https://github.com/ComplexData-MILA/misinfo-datasets -author: Jacob-Junqi Tian -categories: -- Publications -- poli-sci -excerpt: Data is a barrier to reliable misinformation detection solutions. To address this, we curated the largest collection of (mis)information datasets in the literature, totaling 75. From these, we evaluated the quality of all of the 36 datasets that consist of statements or claims. We assessed these datasets to identify those with solid foundations for empirical work and those with flaws that could result in misleading results, such as insufficient label quality, spurious correlations, or political bias. We further show that regardless of label quality, categorical labels may no longer give an accurate evaluation of detection model performance, and discuss ways to solve this. Overall, this guide aims to provide a roadmap for obtaining higher quality data and conducting more effective evaluations in this domain. - - ---- - -*{{ page.names }}* - -**{{ page.venue }}** - -{% include display-publication-links.html pub=page %} - -## Abstract - -Misinformation is a complex societal issue, and mitigating solutions are difficult to create due to data deficiencies. To address this problem, we have curated the largest collection of (mis)information datasets in the literature, totaling 75. From these, we evaluated the quality of all of the 36 datasets that consist of statements or claims. We assess these datasets to identify those with solid foundations for empirical work and those with flaws that could result in misleading and non-generalizable results, such as insufficient label quality, spurious correlations, or political bias. We further provide state-of-the-art baselines on all these datasets, but show that regardless of label quality, categorical labels may no longer give an accurate evaluation of detection model performance. We discuss alternatives to mitigate this problem. Overall, this guide aims to provide a roadmap for obtaining higher quality data and conducting more effective evaluations, ultimately improving research in misinformation detection. All datasets and other artifacts are available at https://misinfo-datasets.complexdatalab.com/. \ No newline at end of file diff --git a/_posts/papers/2025-02-25-10.1609-aaai.v39i27.35005.md b/_posts/papers/2025-02-25-10.1609-aaai.v39i27.35005.md new file mode 100644 index 000000000..5e37b0e50 --- /dev/null +++ b/_posts/papers/2025-02-25-10.1609-aaai.v39i27.35005.md @@ -0,0 +1,24 @@ +--- +title: "Enhancing Privacy in the Early Detection of Sexual Predators Through Federated Learning and Differential Privacy" +venue: Proceedings of the Thirty-ninth AAAI Conference on Artificial Intelligence, AI for Social Impact (AISI) track +names: Khaoula Chehbouni, Martine De Cock, Gilles Caporossi, Afaf Taik, Reihaneh Rabbany, Golnoosh Farnadi +tags: +- AAAI +link: https://doi.org/10.1609/aaai.v39i27.35005 +author: Khaoula Chehbouni +categories: +- Publications +- online-crime +excerpt: Existing approaches to detecting online grooming monitor private conversations through centrally trained models. We implement a privacy-preserving pipeline for the early detection of sexual predators using federated learning and differential privacy, and show on real-world data that privacy and utility can coexist with only a slight reduction in utility. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +The increased screen time and isolation caused by the COVID-19 pandemic have led to a significant surge in cases of online grooming, which is the use of strategies by predators to lure children into sexual exploitation. Previous efforts to detect grooming in industry and academia have involved accessing and monitoring private conversations through centrally-trained models or sending private conversations to a global server. In this work, we implement a privacy-preserving pipeline for the early detection of sexual predators. We leverage federated learning and differential privacy in order to create safer online spaces for children while respecting their privacy. We investigate various privacy-preserving implementations and discuss their benefits and shortcomings. Our extensive evaluation using real-world data proves that privacy and utility can coexist with only a slight reduction in utility. diff --git a/_posts/papers/2025-08-03-10.1145-3711896.3737437.md b/_posts/papers/2025-08-03-10.1145-3711896.3737437.md new file mode 100644 index 000000000..6cfd7d21e --- /dev/null +++ b/_posts/papers/2025-08-03-10.1145-3711896.3737437.md @@ -0,0 +1,29 @@ +--- +title: "A Guide to Misinformation Detection Data and Evaluation" +venue: Proceedings of the 31st SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), Datasets and Benchmarks Track +names: Camille Thibault, Jacob-Junqi Tian, Gabrielle Péloquin-Skulski, Taylor Lynn Curtis, James Zhou, + Florence Laflamme, Yuxiang Guan, Reihaneh Rabbany, Jean-François Godbout, Kellin Pelrine +tags: +- KDD +link: https://dl.acm.org/doi/10.1145/3711896.3737437 +webpage: https://misinfo-datasets.complexdatalab.com/ +code: https://github.com/ComplexData-MILA/misinfo-datasets +award: Best Datasets & Benchmarks Track Paper, Runner-Up +award_url: https://kdd2025.kdd.org/awards/ +author: Camille Thibault +categories: +- Publications +- poli-sci +excerpt: We curated the largest collection of (mis)information datasets in the literature, totaling 75, and evaluated the quality of the 45 claim- and paragraph-based ones, finding that spurious correlations and ambiguous, unverifiable examples affect most of them. We provide state-of-the-art baselines, show that categorical labels may no longer accurately evaluate detection models, and propose Evaluation Quality Assurance (EQA) to guide the field toward better grounded evaluations. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Misinformation is a complex societal issue, and mitigating solutions are difficult to create due to data deficiencies. To address this, we have curated the largest collection of (mis)information datasets in the literature, totaling 75. From these, we evaluated the quality of 36 datasets that consist of statements or claims, as well as the 9 datasets that consist of data in purely paragraph form. We assess these datasets to identify those with solid foundations for empirical work and those with flaws that could result in misleading and non-generalizable results, such as spurious correlations, or examples that are ambiguous or otherwise impossible to assess for veracity. We find the latter issue is particularly severe and affects most datasets in the literature. We further provide state-of-the-art baselines on all these datasets, but show that regardless of label quality, categorical labels may no longer give an accurate evaluation of detection model performance. Finally, we propose and highlight Evaluation Quality Assurance (EQA) as a tool to guide the field toward systemic solutions rather than inadvertently propagating issues in evaluation. Overall, this guide aims to provide a roadmap for higher quality data and better grounded evaluations, ultimately improving research in misinformation detection. All datasets and other artifacts are available at https://misinfo-datasets.complexdatalab.com/. diff --git a/_posts/papers/2025-10-15-10.1609-aies.v8i2.36677.md b/_posts/papers/2025-10-15-10.1609-aies.v8i2.36677.md new file mode 100644 index 000000000..e441811b0 --- /dev/null +++ b/_posts/papers/2025-10-15-10.1609-aies.v8i2.36677.md @@ -0,0 +1,24 @@ +--- +title: "Reframing AI-for-Good: Radical Questioning in AI for Human Trafficking Interventions" +venue: Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society (AIES) +names: Pratheeksha Nair, Gabriel Lefebvre, Maryam Molamohammadi, Sophia Garrel, Reihaneh Rabbany +tags: +- AIES +link: https://doi.org/10.1609/aies.v8i2.36677 +author: Pratheeksha Nair +categories: +- Publications +- online-crime +excerpt: We introduce Radical Questioning, a five-step, pre-design ethics framework for asking whether AI should be applied to a complex social problem, not just how. Developed with survivor-led organizations and piloted in human trafficking, it shifted our design away from automated detection tools and toward survivor-controlled, empowerment-based technologies. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +This paper introduces Radical Questioning (RQ), a structured, pre-design ethics framework developed to assess whether artificial intelligence (AI) should be applied to complex social problems rather than merely how. While much of responsible AI development focuses on aligning systems with principles such as fairness, transparency, and accountability, it often begins after the decision to build has already been made, implicitly treating the deployment of AI as a given rather than a question in itself. In domains such as human trafficking, marked by contested definitions, systemic injustice, and deep stakeholder asymmetries, such assumptions can obscure foundational ethical concerns. RQ offers an upstream, deliberative process for surfacing these concerns before design begins. Drawing from critical theory, participatory ethics, and relational responsibility, RQ formalizes a five-step framework to interrogate problem framings, confront techno-solutionist tendencies, and reflect on the moral legitimacy of intervention. Developed through interdisciplinary collaboration and engagement with survivor-led organizations, RQ was piloted in the domain of human trafficking (HT) which is a particularly high-stakes and ethically entangled application area. Its use led to a fundamental design shift: away from automated detection tools and toward survivor-controlled, empowerment-based technologies. We argue that RQ's novelty lies in both its temporal position, i.e, prior to technical design, and its orientation toward domains where harm is structural and ethical clarity cannot be achieved through one-size-fits-all solutions. RQ thus addresses a critical gap between abstract principles of responsible AI and the lived ethical demands of real-world deployment. diff --git a/_posts/papers/2025-12-02-2406.10426.md b/_posts/papers/2025-12-02-2406.10426.md new file mode 100644 index 000000000..860148320 --- /dev/null +++ b/_posts/papers/2025-12-02-2406.10426.md @@ -0,0 +1,26 @@ +--- +title: "MiNT: Multi-Network Transfer Benchmark for Temporal Graph Learning" +venue: Proceedings of the Thirty-ninth Conference on Neural Information Processing Systems (NeurIPS), Datasets and Benchmarks Track +names: Kiarash Shamsi, Tran Gia Bao Ngo, Razieh Shirzadkhani, Shenyang Huang, Farimah Poursafaei, Poupak Azad, + Reihaneh Rabbany, Baris Coskunuzer, Guillaume Rabusseau, Cuneyt Gurcan Akcora +tags: +- NeurIPS +link: https://papers.neurips.cc/paper_files/paper/2025/hash/c5548168cb7324f714365a971dfe76d1-Abstract-Datasets_and_Benchmarks_Track.html +code: https://github.com/benjaminnNgo/ScalingTGNs +author: Shenyang Huang +categories: +- Publications +- TGL +excerpt: We introduce a benchmark of 84 real-world temporal transaction networks and MiNT, a pre-training framework that captures transferable temporal dynamics across networks. Trained on up to 64 networks and evaluated on 20 unseen ones, MiNT consistently outperforms individually trained models, revealing scaling trends in temporal graph learning and laying the groundwork for temporal graph foundation models. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Temporal Graph Learning (TGL) aims to discover patterns in evolving networks or temporal graphs and leverage these patterns to predict future interactions. However, most existing research focuses on learning from a single network in isolation, leaving the challenges of within-domain and cross-domain generalization largely unaddressed. In this study, we introduce a new benchmark of 84 real-world temporal transaction networks and propose Temporal Multi-network Transfer (MiNT), a pre-training framework designed to capture transferable temporal dynamics across diverse networks. We train MiNT models on up to 64 transaction networks and evaluate their generalization ability on 20 held-out, unseen networks. Our results show that MiNT consistently outperforms individually trained models, revealing a strong relation between the number of pre-training networks and transfer performance. These findings highlight scaling trends in temporal graph learning and underscore the importance of network diversity in improving generalization. This work establishes the first large-scale benchmark for studying transferability in TGL and lays the groundwork for developing Temporal Graph Foundation Models. diff --git a/_posts/papers/2026-04-13-2512.13915.md b/_posts/papers/2026-04-13-2512.13915.md new file mode 100644 index 000000000..2401800ff --- /dev/null +++ b/_posts/papers/2026-04-13-2512.13915.md @@ -0,0 +1,24 @@ +--- +title: "Deepfakes in the 2025 Canadian Election: Prevalence, Partisanship, and Platform Dynamics" +venue: Proceedings of The Web Conference (TheWebConf), Short Paper Track +names: Victor Livernoche, Andreea Musulan, Zachary Yang, Jean-François Godbout, Reihaneh Rabbany +tags: +- TheWebConf +link: https://arxiv.org/abs/2512.13915 +author: Victor Livernoche +categories: +- Publications +- poli-sci +excerpt: Analyzing 187,778 posts from X, Bluesky and Reddit during the 2025 Canadian federal election, we find that 5.86% of election-related images were deepfakes, shared more often by right-leaning accounts. Most were benign or non-political and harmful ones drew little attention, but realistic fabricated images drew higher engagement. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Concerns about AI-generated political content are growing, yet there is limited empirical evidence on how deepfakes actually appear and circulate across social platforms during major events in democratic countries. In this study, we present one of the first in-depth analyses of how these realistic synthetic media shape the political landscape online, focusing specifically on the 2025 Canadian federal election. By analyzing 187,778 posts from X, Bluesky, and Reddit with a high-accuracy detection framework trained on a diverse set of modern generative models, we find that 5.86% of election-related images were deepfakes. Right-leaning accounts shared them more frequently, with 8.66% of their posted images flagged compared to 4.42% for left-leaning users, often with defamatory or conspiratorial intent. Yet, most detected deepfakes were benign or non-political, and harmful ones drew little attention, accounting for only 0.12% of all views on X. Overall, deepfakes were present in the election conversation, but their reach was modest, and realistic fabricated images, although less common, drew higher engagement, highlighting growing concerns about their potential misuse. diff --git a/_posts/papers/2026-04-23-2510.07586.md b/_posts/papers/2026-04-23-2510.07586.md new file mode 100644 index 000000000..b83ce3975 --- /dev/null +++ b/_posts/papers/2026-04-23-2510.07586.md @@ -0,0 +1,27 @@ +--- +title: "TGM: A Modular and Efficient Library for Machine Learning on Temporal Graphs" +venue: Proceedings of the Fourteenth International Conference on Learning Representations (ICLR) +names: Jacob Chmura, Shenyang Huang, Tran Gia Bao Ngo, Ali Parviz, Farimah Poursafaei, Jure Leskovec, + Michael M. Bronstein, Guillaume Rabusseau, Matthias Fey, Reihaneh Rabbany +tags: +- ICLR +link: https://arxiv.org/abs/2510.07586 +code: https://github.com/tgm-team/tgm +webpage: https://tgm.readthedocs.io +author: Shenyang Huang +categories: +- Publications +- TGL +excerpt: We introduce TGM, a research-oriented library for machine learning on temporal graphs and the first to unify continuous- and discrete-time dynamic graph methods. It supports dynamic node features, time-granularity conversions and link-, node- and graph-level tasks, with an average 7.8x speedup over DyGLib and 175x on graph discretization. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Well-designed open-source software drives progress in Machine Learning (ML) research. While static graph ML enjoys mature frameworks like PyTorch Geometric and DGL, ML for temporal graphs (TG), networks that evolve over time, lacks comparable infrastructure. Existing TG libraries are often tailored to specific architectures, hindering support for diverse models in this rapidly evolving field. Additionally, the divide between continuous- and discrete-time dynamic graph methods (CTDG and DTDG) limits direct comparisons and idea transfer. To address these gaps, we introduce Temporal Graph Modelling (TGM), a research-oriented library for ML on temporal graphs, the first to unify CTDG and DTDG approaches. TGM offers first-class support for dynamic node features, time-granularity conversions, and native handling of link-, node-, and graph-level tasks. Empirically, TGM achieves an average 7.8x speedup across multiple models, datasets, and tasks compared to the widely used DyGLib, and an average 175x speedup on graph discretization relative to available implementations. Beyond efficiency, we show in our experiments how TGM unlocks entirely new research possibilities by enabling dynamic graph property prediction and time-driven training paradigms, opening the door to questions previously impractical to study. TGM is available at https://github.com/tgm-team/tgm. diff --git a/_posts/papers/2026-04-23-2511.07332.md b/_posts/papers/2026-04-23-2511.07332.md new file mode 100644 index 000000000..f756812a8 --- /dev/null +++ b/_posts/papers/2026-04-23-2511.07332.md @@ -0,0 +1,24 @@ +--- +title: "Grounding Computer Use Agents on Human Demonstrations" +venue: Proceedings of the Fourteenth International Conference on Learning Representations (ICLR) +names: Aarash Feizi, Shravan Nayak, Xiangru Jian, Kevin Qinghong Lin, Kaixin Li, Rabiul Awal, Xing Han Lù, + Johan Obando-Ceron, Juan A. Rodriguez, Nicolas Chapados, David Vazquez, Adriana Romero-Soriano, + Reihaneh Rabbany, Perouz Taslakian, Christopher Pal, Spandana Gella, Sai Rajeswar +tags: +- ICLR +link: https://arxiv.org/abs/2511.07332 +author: Aarash Feizi +categories: Publications +excerpt: We introduce GroundCUA, a large-scale desktop grounding dataset built from expert human demonstrations, covering 87 applications and over 3.56M human-verified annotations. The GroundNext models trained on it achieve state-of-the-art grounding across five benchmarks with less than one-tenth the training data of prior work. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Building reliable computer-use agents requires grounding: accurately connecting natural language instructions to the correct on-screen elements. While large datasets exist for web and mobile interactions, high-quality resources for desktop environments are limited. To address this gap, we introduce GroundCUA, a large-scale desktop grounding dataset built from expert human demonstrations. It covers 87 applications across 12 categories and includes 56K screenshots, with every on-screen element carefully annotated for a total of over 3.56M human-verified annotations. From these demonstrations, we generate diverse instructions that capture a wide range of real-world tasks, providing high-quality data for model training. Using GroundCUA, we develop the GroundNext family of models that map instructions to their target UI elements. At both 3B and 7B scales, GroundNext achieves state-of-the-art results across five benchmarks using supervised fine-tuning, while requiring less than one-tenth the training data of prior work. Reinforcement learning post-training further improves performance, and when evaluated in an agentic setting on the OSWorld benchmark using o3 as planner, GroundNext attains comparable or superior results to models trained with substantially more data. These results demonstrate the critical role of high-quality, expert-driven datasets in advancing general-purpose computer-use agents. diff --git a/_posts/papers/2026-05-25-pmlr-v318-yang26a.md b/_posts/papers/2026-05-25-pmlr-v318-yang26a.md new file mode 100644 index 000000000..2497cb889 --- /dev/null +++ b/_posts/papers/2026-05-25-pmlr-v318-yang26a.md @@ -0,0 +1,24 @@ +--- +title: "ToxiSight: Leveraging Moderator Expertise Through Behavioral Measurement in Gaming Toxicity Annotation" +venue: Proceedings of the Canadian Conference on Artificial Intelligence (Canadian AI) +names: Zachary Yang, Vicki Chen, Domenico Tullo, Reihaneh Rabbany +tags: +- Canadian AI +link: https://proceedings.mlr.press/v318/yang26a.html +author: Zachary Yang +categories: +- Publications +- online-toxicity +excerpt: We present ToxiSight, an annotation platform that treats moderators' hesitation, revision and disagreement as signals of genuinely ambiguous content and failing guidelines, rather than resolving them into consensus. Deployed with 10 professional moderators across 60,000 lines of gaming chat, it exposed systematic category failures invisible to inter-annotator metrics and raised completion rates from 60% to 95%. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Content moderation systems commonly treat human annotators as interchangeable label sources, resolving disagreements through majority voting or expert arbitration. We present ToxiSight, an annotation platform that reframes this assumption: rather than extracting consensus, the system supports moderator reasoning by treating hesitation, revision, and disagreement as signals revealing where content is genuinely ambiguous and where taxonomic guidelines fail. ToxiSight integrates gaming-specific contextual widgets with behavioral telemetry, capturing the cognitive processes underlying toxicity validation decisions. Through deployment with 10 professional moderators across 60,000 lines of gaming chat, we demonstrate that behavioral patterns expose systematic category failures invisible to traditional inter-annotator metrics. The Controversial category shows 72% revision rates with fast processing times, indicating immediate recognition of definitional breakdown, while Threats (Life-Threatening) exhibits 75% revisions with slow processing, signaling genuine interpretive complexity. Completion rates improved from 60% to 95%, and moderators reported reduced decision stress when permitted to express uncertainty. This case study demonstrates that trustworthy toxicity detection requires annotation systems designed around the irreducible complexity of human judgment, not against it. diff --git a/_posts/papers/2026-06-01-ssrn.6867358.md b/_posts/papers/2026-06-01-ssrn.6867358.md new file mode 100644 index 000000000..fa3c06400 --- /dev/null +++ b/_posts/papers/2026-06-01-ssrn.6867358.md @@ -0,0 +1,28 @@ +--- +title: "AI Epistemic Risks: Emerging Mechanisms & Evidence" +venue: SSRN +names: Stephen Casper, Jonathan Stray, Anna Gausen, Cameron Jones, Brian Christian, Bálint Gyevnár, + Siao Si Looi, Joshua Levy, Elizabeth Seger, Hannah Kirk, Kobi Hackenburg, Matt Kowal, Michelle Malonza, + Dan Zhao, Zhonghao He, Luke Hewitt, Hause Lin, Maarten Sap, Dylan Hadfield-Menell, Atoosa Kasirzadeh, + Gordon Pennycook, Jean-François Godbout, Kellin Pelrine, Reihaneh Rabbany, Natasha Jaques, Jasmine Li, + Yoshua Bengio, David Rand, Thomas Costello, Mick Yang +tags: +- SSRN +link: https://ssrn.com/abstract=6867358 +author: Jean-Francois Godbout +categories: +- Publications +- poli-sci +excerpt: AI poses risks to humanity's collective capacity to form accurate beliefs, reason well and maintain a healthy information environment. We identify three mechanisms of systemic epistemic decline (persuasion and manipulation, cognitive offloading, and feedback loops), record the emerging empirical evidence for each, and analyze how they amplify one another. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Advances in artificial intelligence pose risks to humanity's collective capacity to form accurate beliefs, reason well, and maintain a healthy information environment: risks we term epistemic risks. These risks arise from AI's integration into the infrastructure through which individuals and societies think, form beliefs, and make sense of the world together. We highlight three primary mechanisms by which AI can cause systemic epistemic decline: persuasion and manipulation, cognitive offloading, and feedback loops. AI can persuade and manipulate through misuse for economic or political manipulation, inciting crime or radicalization, and escalating conflict, and can also cause unintentional harms such as sycophancy and mental health risks. We record the emerging empirical evidence for each mechanism and analyze how they likely amplify one another, creating compounding systemic risk. The evidence suggests that AI could be an unprecedented lever for improving epistemics, but this will not happen by default. diff --git a/_posts/papers/2026-07-06-icml2026-validation-gap.md b/_posts/papers/2026-07-06-icml2026-validation-gap.md new file mode 100644 index 000000000..532ad0686 --- /dev/null +++ b/_posts/papers/2026-07-06-icml2026-validation-gap.md @@ -0,0 +1,18 @@ +--- +title: "Position: Time to Close The Validation Gap in LLM Social Simulations" +venue: Proceedings of the Forty-third International Conference on Machine Learning (ICML), Position Paper Track +names: Maximilian Puelma Touzel, Sneheel Sarangi, Aurélien Bück-Kaeffer, Zachary Yang, Jean-François Godbout, Reihaneh Rabbany +tags: +- ICML +link: https://openreview.net/forum?id=LpbxLBcOBf +author: Maximilian Puelma Touzel +categories: Publications +excerpt: LLM-based social simulations are rapidly proliferating across policy analysis, epidemiology and computational social science, yet the field lacks consensus on how to validate them. We argue for shifting from expansion to consolidation, through shared benchmarks, open data and reproducible evaluation grounded in social science, and outline a concrete research program toward that goal. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} diff --git a/_posts/papers/2026-08-09-10.1145-3770855.3817476.md b/_posts/papers/2026-08-09-10.1145-3770855.3817476.md new file mode 100644 index 000000000..a3460c84c --- /dev/null +++ b/_posts/papers/2026-08-09-10.1145-3770855.3817476.md @@ -0,0 +1,25 @@ +--- +title: "CrediBench: Web-Scale Network Data for Credibility Prediction" +venue: Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V. 2 +names: Emma Kondrup, Sebastian Sabry, Hussein Abdallah, Zachary Yang, Jiaqi Xiong, Kellin Pelrine, + James Zhou, Zhijin Guo, Michael M Bronstein, Jean-François Godbout, Reihaneh Rabbany, Shenyang Huang +tags: +- KDD +link: https://dl.acm.org/doi/pdf/10.1145/3770855.3817476 +author: Emma Kondrup +categories: +- Publications +- poli-sci +excerpt: We present CrediBench, a web-scale dataset of temporal web graphs built from Common Crawl, with over 40 million nodes, their webpage content and over 1 billion hyperlink edges per monthly snapshot, plus credibility labels for 662,575 web domains. Combining graph, text and time signals substantially improves web domain credibility prediction over existing baselines. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Automatically assessing the credibility of online sources presents an invaluable tool for navigating today's information ecosystem. However, existing approaches either depend on scarce and costly human annotations, or focus exclusively on assessments at the level of individual claims. Misinformation often spreads via interlinked web domains, whose connections evolve over time. Focusing on claims alone ignores these structural and temporal credibility signals evident in the changing web topology. Existing datasets fail to capture these central modalities in web domain credibility prediction: namely, internet topology, temporality and text (webpage) content. To address this gap, we present CrediBench, a dataset containing eight months of web graph data; of which we analyze the three months surrounding the 2024 U.S. federal elections, a time of heightened misinformation propagation online. Each monthly snapshot contains over 40 million nodes, their scraped webpage content, and over 1 billion hyperlink edges. CrediBench supports credibility prediction as both a regression (continuous credibility score) and a binary classification task (credible or not). For classification, we curate a novel binary label set containing 662,575 web domains labelled for boolean credibility, spanning four areas (misinformation, crowd-sourced, malware and phishing). Our empirical experiments support that all task modalities (graph, text and time) contribute significantly to achieving the best performance. In particular, our multi-modal regression model trained on CrediBench outperforms other configurations and existing baselines, decreasing Mean Average Error from 0.162 to 0.107 on the regression task, while the multi-modal classifier improves accuracy from 56% to 85% on the classification one. CrediBench, our proposed web-scale multi-modal dataset, is available on Huggingface for future research. diff --git a/_posts/papers/2026-08-15-rlc2026-dart.md b/_posts/papers/2026-08-15-rlc2026-dart.md new file mode 100644 index 000000000..b73196815 --- /dev/null +++ b/_posts/papers/2026-08-15-rlc2026-dart.md @@ -0,0 +1,22 @@ +--- +title: "DART: Dual Adaptive Residual Tracking for Low-Bias Advantage Estimation and Credit Assignment in AI Agents" +venue: Proceedings of the Reinforcement Learning Conference (RLC) +names: Shahrad Mohammadzadeh, Amir-massoud Farahmand, Reihaneh Rabbany, Doina Precup +tags: +- RLC +link: https://rlj.cs.umass.edu/2026/papers/Paper74.html +author: Shahrad Mohammadzadeh +categories: Publications +excerpt: We show that the bottleneck of Generalized Advantage Estimation is dissemination bias, which can be corrected without discarding the critic. DART pairs a stable base critic with a residual critic that tracks a Monte Carlo correction, provably lowering steady-state error and matching or outperforming PPO across dm_control, especially on sparse and long-horizon tasks. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Proximal Policy Optimization (PPO) is the primary workhorse for complex continuous control and Reinforcement Learning from Human Feedback (RLHF). However, the reliability of Generalized Advantage Estimation (GAE) has recently been called into question: critic-free methods like RLOO and GRPO have raised the question of whether value-induced bias is doing more harm than the variance reduction it provides. We show this is the wrong dichotomy. The bottleneck is the dissemination bias of GAE, where value approximation error propagates into advantage estimates and misleads the policy gradient, a bias that can be corrected without discarding the critic. We propose Dual Adaptive Residual Tracking (DART), a versatile two-critic scheme that decouples the bias–variance trade-off. DART pairs a stable base critic learning from low-variance bootstrapped targets with a residual critic that tracks a high-fidelity Monte Carlo correction via a frozen-target mechanism. Theoretically, under linear function approximation assumptions, we prove that DART achieves a strictly smaller steady-state mean-squared error (MSE) upper bound than single-critic methods in high-variance regimes. Empirically, across all dm_control continuous control environments, DART matches or outperforms PPO in aggregate, with clear advantages in sparse and long-horizon tasks. We further provide initial evidence of DART's effectiveness in LLM fine-tuning. diff --git a/_posts/papers/2026-09-08-10.5753-sbbd_estendido.2026.249933.md b/_posts/papers/2026-09-08-10.5753-sbbd_estendido.2026.249933.md new file mode 100644 index 000000000..d622067a3 --- /dev/null +++ b/_posts/papers/2026-09-08-10.5753-sbbd_estendido.2026.249933.md @@ -0,0 +1,25 @@ +--- +title: "SynTraffic: A synthetic dataset for spatiotemporal human trafficking detection in heterogeneous graphs" +venue: Simpósio Brasileiro de Banco de Dados (SBBD) +names: Joao Lopes, Ana Nascimento, Pratheeksha Nair, Catalina Vajiac, Andreas Olligschlaeger, Reihaneh Rabbany, + Christos Faloutsos, Mirela Cazzolato +tags: +- SBBD +link: https://doi.org/10.5753/sbbd_estendido.2026.249933 +author: Pratheeksha Nair +categories: +- Publications +- online-crime +excerpt: Human trafficking advertisement data is sensitive and hard to share, which limits research on automated detection. We present SynTraffic, a privacy-safe synthetic dataset and reproducible generation methodology that statistically reproduces a real corpus' posting volume, spatial mobility, temporal dynamics and metadata-sharing structure. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Online escort platforms have become a common vector for human trafficking (HT) networks to advertise victims alongside legitimate listings, making manual screening infeasible. Progress on automated analysis is limited by restricted data access: HT advertisement corpora carry personal, potentially victim-identifying information, imposing ethical and legal constraints on sharing and reproducibility. This work addresses that gap with a reproducible methodology for generating synthetic HT advertisement data that is statistically representative of a real corpus, without exposing sensitive content. The proposed dataset SynTraffic reproduces posting volume, spatial mobility, temporal dynamics, and metadata-sharing structure, including the size distribution of identity clusters formed by shared contact metadata. We statistically validate SynTraffic testing distributional fidelity and cluster-size fidelity. SynTraffic is openly available, as well as the generation methodology, presenting a privacy-safe substitute for the real corpus. diff --git a/_posts/papers/2026-11-01-2605.00513.md b/_posts/papers/2026-11-01-2605.00513.md new file mode 100644 index 000000000..90807ac86 --- /dev/null +++ b/_posts/papers/2026-11-01-2605.00513.md @@ -0,0 +1,22 @@ +--- +title: "ControBench: An Interaction-Aware Benchmark for Controversial Discourse Analysis on Social Networks" +venue: The IEEE International Conference on Data Mining (ICDM) +names: Thanh Thuy Ta, Jiaqi Zhu, Xuan Liu, Shang Lin, Reihaneh Rabbany, Guillaume Rabusseau, Lihui Chen, Yilun Zheng, Sitao Luan +tags: +- ICDM +link: https://arxiv.org/abs/2605.00513 +author: Sitao Luan +categories: Publications +excerpt: We introduce ControBench, a benchmark for controversial discourse analysis that combines heterogeneous social interaction graphs with rich textual semantics, built from Reddit discussions on Trump, abortion and religion (7,370 users, 1,783 posts, 26,525 interactions). Its low or negative homophily reflects the cross-cutting structure of real debate, making it a challenging testbed for graph neural networks and language models. + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} + +## Abstract + +Understanding how people argue across ideological divides online is important for studying political polarization, misinformation, and content moderation. Existing datasets capture only part of this problem: some preserve text but ignore interaction structure, some model structure without rich semantics, and others represent conversations without stable user-level ideological identity. We introduce ControBench, a benchmark for controversial discourse analysis that combines heterogeneous social interaction graphs with rich textual semantics. Built from Reddit discussions on three topics, Trump, abortion, and religion, ControBench contains 7,370 users, 1,783 posts, and 26,525 interactions. The graph contains user and post nodes connected by semantically enriched edges; in particular, user-comment-user edges encode both a reply and the parent comment that it responds to, preserving local argumentative context. User labels are derived from self-declared Reddit flairs, providing a scalable proxy for ideological identity without manual annotation. The resulting datasets exhibit low or negative adjusted homophily (Trump: -0.77, Abortion: 0.06, Religion: 0.04), reflecting the cross-cutting structure of real-world debate. We evaluate graph neural networks, pretrained language models, and large language models on ControBench and observe distinct performance patterns across topics and model families, especially when ideological boundaries are ambiguous. These results position ControBench as a challenging and realistic benchmark for controversial discourse analysis. diff --git a/_posts/papers/2026-11-01-emnlp2026-simulate-users.md b/_posts/papers/2026-11-01-emnlp2026-simulate-users.md new file mode 100644 index 000000000..7b76122db --- /dev/null +++ b/_posts/papers/2026-11-01-emnlp2026-simulate-users.md @@ -0,0 +1,16 @@ +--- +title: "Can LLMs Learn to Simulate Users from Behavioral History? An Empirical Study" +venue: The 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP) +names: Sneheel Sarangi, Aurélien Bück-Kaeffer, Maximilian Puelma Touzel, Zachary Yang, Jean-François Godbout, Reihaneh Rabbany +tags: +- EMNLP +author: Sneheel Sarangi +categories: Publications + +--- + +*{{ page.names }}* + +**{{ page.venue }}** + +{% include display-publication-links.html pub=page %} diff --git a/_posts/teaching/2019-09-01-comp596-F19.md b/_posts/teaching/2019-09-01-comp599-F19.md similarity index 94% rename from _posts/teaching/2019-09-01-comp596-F19.md rename to _posts/teaching/2019-09-01-comp599-F19.md index 1d08a4a40..0d99f5ad4 100644 --- a/_posts/teaching/2019-09-01-comp596-F19.md +++ b/_posts/teaching/2019-09-01-comp599-F19.md @@ -1,7 +1,7 @@ --- title: Network Science term: Fall 2019 -course_code: COMP 596 +course_code: COMP 599 author: Reihaneh Rabbany names: Reihaneh Rabbany categories: Teaching @@ -16,7 +16,7 @@ tags: - Network Science --- -* **Course codes:** COMP 596 (Fall 2019) +* **Course codes:** COMP 599 (Fall 2019) * **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) * **Classroom:** ENGMC 103 * **Time:** Tuesdays and Thursdays, 10:00 am – 11:30 pm diff --git a/_posts/teaching/2020-01-01-comp551-W20.md b/_posts/teaching/2020-01-01-comp551-W20.md index a8f8cf24b..1ce37063b 100644 --- a/_posts/teaching/2020-01-01-comp551-W20.md +++ b/_posts/teaching/2020-01-01-comp551-W20.md @@ -3,7 +3,7 @@ title: Applied Machine Learning term: Winter 2020 course_code: COMP 551 author: Reihaneh Rabbany -names: Reihaneh Rabbany +names: Reihaneh Rabbany w/ Siamak Ravanbakhsh categories: Teaching share: false last_modified_at: 2024-10-11T00:00:00 @@ -17,7 +17,7 @@ tags: --- * **Course codes:** COMP 551 (Winter 2020) -* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) +* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/), Siamak Ravanbakhsh * **Classroom:** Strathcona Anatomy & Dentistry M-1 * **Time:** Tuesdays and Thursdays, 11:30 am - 12:55 am * **Course Website:** [here](http://www.reirab.com/Teaching/AML20/index.html) diff --git a/_posts/teaching/2020-09-01-comp596-F20.md b/_posts/teaching/2020-09-01-comp599-F20.md similarity index 95% rename from _posts/teaching/2020-09-01-comp596-F20.md rename to _posts/teaching/2020-09-01-comp599-F20.md index f7d597a05..f3acf9fbb 100644 --- a/_posts/teaching/2020-09-01-comp596-F20.md +++ b/_posts/teaching/2020-09-01-comp599-F20.md @@ -1,7 +1,7 @@ --- title: Network Science term: Fall 2020 -course_code: COMP 596 +course_code: COMP 599 author: Reihaneh Rabbany names: Reihaneh Rabbany categories: Teaching @@ -16,7 +16,7 @@ tags: - Network Science --- -* **Course codes:** COMP 596 (Fall 2020) +* **Course codes:** COMP 599 (Fall 2020) * **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) * **Location:** Remote through MyCourses * **Time:** Monday & Wednesday, 10:00 am – 11:30 pm diff --git a/_posts/teaching/2024-09-01-comp551-F24.md b/_posts/teaching/2024-09-01-comp551-F24.md index 695788596..d22bd04ab 100644 --- a/_posts/teaching/2024-09-01-comp551-F24.md +++ b/_posts/teaching/2024-09-01-comp551-F24.md @@ -1,9 +1,9 @@ --- title: Applied Machine Learning -term: Winter 2025 +term: Fall 2024 course_code: COMP 551 author: Reihaneh Rabbany -names: Reihaneh Rabbany +names: Reihaneh Rabbany w/ Isabeau Prémont-Schwarz categories: Teaching share: false last_modified_at: 2024-10-11T00:00:00 @@ -12,11 +12,11 @@ toc: true toc_sticky: true toc_label: "Table of Contents" tags: - - Winter 2025 + - Fall 2024 - Applied Machine Learning --- -* **Course codes:** COMP 551 (Winter 2025) +* **Course codes:** COMP 551 (Fall 2024) * **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/), [Isabeau Prémont-Schwarz](https://www.cs.mcgill.ca/~isabeau/) * **Location:** McConnell 204 * **Time:** Mondays and Wednesdays, 14:35-15:55 diff --git a/_posts/teaching/2025-01-01-comp511-W25.md b/_posts/teaching/2025-01-01-comp511-W25.md new file mode 100644 index 000000000..0ddcfc25b --- /dev/null +++ b/_posts/teaching/2025-01-01-comp511-W25.md @@ -0,0 +1,24 @@ +--- +title: Network Science +term: Winter 2025 +course_code: COMP 511 +author: Reihaneh Rabbany +names: Reihaneh Rabbany +categories: Teaching +share: false +last_modified_at: 2026-10-12T00:00:00 +layout: single +toc: true +toc_sticky: true +toc_label: "Table of Contents" +tags: + - Winter 2025 + - Network Science +--- + +* **Course codes:** COMP 511 (Winter 2025) +* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) + +# Overview + +This is on Network Science, Graph Mining and Graph Learning. Networks model relationships in complex systems, from hyperlinks between webpages, and co-authorships between research scholars to biological interactions between proteins and genes, and synaptic links between neurons. Network Science is an interdisciplinary research area involving researchers from Physics, Computer Science, Sociology, Math and Statistics, with applications in a wide range of domains including Biology, Medicine, Political Science, Marketing, Ecology, Criminology, etc. In this course, we will cover the basic concepts and techniques used in Network Science, review the state of the art techniques, and discuss the most recent developments. diff --git a/_posts/teaching/2025-01-01-comp551-W25.md b/_posts/teaching/2025-09-01-comp551-F25.md similarity index 63% rename from _posts/teaching/2025-01-01-comp551-W25.md rename to _posts/teaching/2025-09-01-comp551-F25.md index 93214a864..6847d6d29 100644 --- a/_posts/teaching/2025-01-01-comp551-W25.md +++ b/_posts/teaching/2025-09-01-comp551-F25.md @@ -1,26 +1,23 @@ --- -title: Applied Machine Learning -term: Winter 2025 +title: Applied Machine Learning +term: Fall 2025 course_code: COMP 551 author: Reihaneh Rabbany -names: Reihaneh Rabbany +names: Reihaneh Rabbany w/ Sékou-Oumar Kaba categories: Teaching share: false -last_modified_at: 2024-10-11T00:00:00 +last_modified_at: 2026-10-12T00:00:00 layout: single toc: true toc_sticky: true toc_label: "Table of Contents" tags: - - Winter 2025 + - Fall 2025 - Applied Machine Learning --- -* **Course codes:** COMP 551 (Winter 2025) -* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) -* **Location:** Stewart Biology Building S1/4 (lectures will be recorded) -* **Time:** Tuesdays and Thursdays, 1:05 pm - 2:25 pm -* **Course Website:** [here](http://www.reirab.com/Teaching/AML23/index.html) +* **Course codes:** COMP 551 (Fall 2025) +* **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/), Sékou-Oumar Kaba # Overview diff --git a/_posts/teaching/2024-09-01-comp599-F24.md b/_posts/teaching/2026-01-01-comp511-W26.md similarity index 81% rename from _posts/teaching/2024-09-01-comp599-F24.md rename to _posts/teaching/2026-01-01-comp511-W26.md index 9c0a7a2ad..62269eadd 100644 --- a/_posts/teaching/2024-09-01-comp599-F24.md +++ b/_posts/teaching/2026-01-01-comp511-W26.md @@ -1,27 +1,25 @@ --- title: Network Science -term: Fall 2024 -course_code: COMP 599 +term: Winter 2026 +course_code: COMP 511 author: Reihaneh Rabbany names: Reihaneh Rabbany categories: Teaching share: false -last_modified_at: 2024-10-11T00:00:00 +last_modified_at: 2026-10-12T00:00:00 layout: single toc: true toc_sticky: true toc_label: "Table of Contents" tags: - - Fall 2024 + - Winter 2026 - Network Science --- -* **Course codes:** COMP 599 (Fall 2024) +* **Course codes:** COMP 511 (Winter 2026) * **Instructors:** [Reihaneh Rabbany](http://www.reirab.com/) # Overview This is on Network Science, Graph Mining and Graph Learning. Networks model relationships in complex systems, from hyperlinks between webpages, and co-authorships between research scholars to biological interactions between proteins and genes, and synaptic links between neurons. Network Science is an interdisciplinary research area involving researchers from Physics, Computer Science, Sociology, Math and Statistics, with applications in a wide range of domains including Biology, Medicine, Political Science, Marketing, Ecology, Criminology, etc. In this course, we will cover the basic concepts and techniques used in Network Science, review the state of the art techniques, and discuss the most recent developments. - -**Please note that about half of the class is paper discussions and student presentations.** \ No newline at end of file diff --git a/_research_directions/online-crime.md b/_research_directions/online-crime.md index c1b445f3b..8a7971aa3 100644 --- a/_research_directions/online-crime.md +++ b/_research_directions/online-crime.md @@ -18,6 +18,8 @@ logos: name: NSERC - image: /assets/images/logo/SAMSUNG.png name: SAMSUNG + - image: /assets/images/logo/IVADO.png + name: IVADO --- @@ -33,6 +35,6 @@ Sex trafficking impacts 4.8 million people globally and is a $99 billion USD ind # Funding -We acknowledge funding from the Canadian Institute for Advanced Research (CIFAR AI Chair Program), Mila - Quebec AI Institute, Natural Sciences and Engineering Research Council of Canada (NSERC), and Samsung. +We acknowledge funding from the Canadian Institute for Advanced Research (CIFAR AI Chair Program), Mila - Quebec AI Institute, Natural Sciences and Engineering Research Council of Canada (NSERC), Samsung, and IVADO (Institute for Data Valorization). {% include logo-grid.html id="logos" %} diff --git a/_research_directions/poli-sci.md b/_research_directions/poli-sci.md index 65aa26c2a..13ac954f4 100644 --- a/_research_directions/poli-sci.md +++ b/_research_directions/poli-sci.md @@ -24,6 +24,8 @@ logos: name: UKRI - image: /assets/images/logo/IVADO.png name: IVADO + - image: /assets/images/logo/AISI.png + name: AISI --- The current digital age fundamentally changes how we as individuals and a society collect and distribute information. With these changes come powerful new approaches to influence the public agenda. However, the quality of shared information varies widely, and can have wide-reaching consequences. As the sophistication and accessibility of AI tools continue to expand, the authenticity of digital content has been increasingly called into question. Our team stands at the forefront of the challenge to safeguard digital space, as we assess the nature of content and how it proliferates online. For more information check out our [a4.institute](https://www.ai4.institute/). @@ -38,6 +40,6 @@ The current digital age fundamentally changes how we as individuals and a societ # Funding -We acknowledge funding from the Canadian Institute for Advanced Research (CIFAR AI Chair Program), the Fonds de recherche du Québec – Société et culture (FRQSC), the Social Sciences and Humanities Research Council (SSHRC), Mila - Quebec AI Institute,the Government of Canada, UK Research and Innovation (UKRI), and IVADO (Institute for Data Valorization). +We acknowledge funding from the Canadian Institute for Advanced Research (CIFAR AI Chair Program), the Fonds de recherche du Québec – Société et culture (FRQSC), the Social Sciences and Humanities Research Council (SSHRC), Mila - Quebec AI Institute,the Government of Canada, UK Research and Innovation (UKRI), IVADO (Institute for Data Valorization), and the UK AI Security Institute (AISI). {% include logo-grid.html id="logos" %} \ No newline at end of file diff --git a/_sass/custom/display-highlighted-publications.scss b/_sass/custom/display-highlighted-publications.scss index 42486d2f0..89c5b82a5 100644 --- a/_sass/custom/display-highlighted-publications.scss +++ b/_sass/custom/display-highlighted-publications.scss @@ -1,30 +1,22 @@ +/* Selected publications on research direction pages */ + article.highlighted-publication-card { - display: grid; - grid-template-areas: - "title title" - "excerpt excerpt" - "venue venue" - "links links"; - grid-template-columns: 2fr 1fr; - column-gap: 15px; - margin-top: 1.5em; - align-items: start; - align-content: start; - - @media (max-width: 600px) { - grid-template-areas: - "title" - "excerpt" - "venue" - "links"; - grid-template-columns: 1fr; - } + display: flex; + flex-direction: column; + gap: 0.5rem; + padding: 1.25rem 1.5rem; + background: #fff; + border: 1px solid rgba(0, 0, 0, 0.07); + border-radius: 12px; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04); } .highlighted-publication-title { - grid-area: title; - margin: 0 0 5px 0; + margin: 0; + padding: 0; border-bottom: none !important; + font-size: 1.05rem; + line-height: 1.35; >a { text-decoration: none; @@ -32,9 +24,9 @@ article.highlighted-publication-card { } .highlighted-publication-excerpt { - grid-area: excerpt; - display: block; - font-style: italic; + font-size: 0.8rem; + line-height: 1.6; + color: #555; >p { margin: 0; @@ -42,12 +34,13 @@ article.highlighted-publication-card { } .highlighted-publication-venue { - grid-area: venue; - font-weight: bold; + font-size: 0.75rem; + font-weight: 600; + color: #333; } -.publication-links { - grid-area: links; +article.highlighted-publication-card .publication-links { + font-size: 0.75rem; } .publication-ext-link { @@ -57,5 +50,78 @@ article.highlighted-publication-card { } .highlighted-publications-spacing { - margin-bottom: 1rem; // Adjust this value to set the desired spacing -} \ No newline at end of file + height: 1rem; +} + +/* "Show N more" toggle under the first few selected publications (posts-highlighted-publications.html): + a floating pill centred on a hairline divider. */ +.highlighted-publications__toggle-row { + display: flex; + align-items: center; + gap: 1rem; + margin: 0.75rem 0 0; + + &::before, + &::after { + content: ""; + flex: 1; + height: 1px; + background: linear-gradient(to right, transparent, rgba(0, 0, 0, 0.1)); + } + + &::after { + background: linear-gradient(to left, transparent, rgba(0, 0, 0, 0.1)); + } +} + +.highlighted-publications__toggle { + display: inline-flex; + align-items: center; + gap: 0.6rem; + margin: 0; + padding: 0.4rem 0.45rem 0.4rem 1.1rem; + border: 1px solid rgba(0, 0, 0, 0.08); + border-radius: 999px; + background: #fff; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04), 0 4px 14px rgba(0, 0, 0, 0.06); + color: #333; + font-family: inherit; + font-size: 0.8rem; + font-weight: 600; + letter-spacing: 0.01em; + cursor: pointer; + transition: box-shadow 0.2s ease, transform 0.2s ease, color 0.2s ease; + + &:hover, + &:focus-visible { + color: $link-color; + transform: translateY(-1px); + box-shadow: 0 2px 4px rgba(0, 0, 0, 0.05), 0 8px 22px rgba(0, 0, 0, 0.09); + } + + &:focus-visible { + outline: 2px solid rgba($link-color, 0.35); + outline-offset: 2px; + } +} + +.highlighted-publications__toggle-icon { + display: inline-flex; + align-items: center; + justify-content: center; + width: 1.6rem; + height: 1.6rem; + border-radius: 50%; + background: rgba($link-color, 0.1); + color: $link-color; + font-size: 0.65rem; + transition: transform 0.25s ease, background 0.2s ease; + + .highlighted-publications__toggle:hover & { + background: rgba($link-color, 0.16); + } + + .highlighted-publications__toggle[aria-expanded="true"] & { + transform: rotate(180deg); + } +} diff --git a/_sass/custom/display-publications.scss b/_sass/custom/display-publications.scss index daff0e085..5d32e78c0 100644 --- a/_sass/custom/display-publications.scss +++ b/_sass/custom/display-publications.scss @@ -1,31 +1,38 @@ +/* Mirrors the selected-publication cards on research pages (display-highlighted-publications.scss), + with the author list in place of the summary. */ article.publication-card { - display: grid; - grid-template-areas: - "thumbnail title" - "thumbnail authors" - "thumbnail venue" - "thumbnail links"; - grid-template-columns: 300px 1fr; - gap: 10px; - margin: 0.75em 0; - padding: 1em; + display: flex; + flex-direction: column; + gap: 0.5rem; + margin: 0 0 1rem; + padding: 1.25rem 1.5rem; background: #fff; - border-radius: 4px; - box-shadow: 0 1px 3px rgba(0, 0, 0, 0.12); - transition: box-shadow 0.3s ease; + border: 1px solid rgba(0, 0, 0, 0.07); + border-radius: 12px; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04); - &:hover { - box-shadow: 0 3px 6px rgba(0, 0, 0, 0.16); - } - - @media (max-width: 600px) { + // Only reserve an image column for papers that set `thumbnail` in front matter. + &--with-thumbnail { + display: grid; grid-template-areas: - "title" - "authors" - "venue" - "links" - "thumbnail"; - grid-template-columns: 1fr; + "thumbnail title" + "thumbnail authors" + "thumbnail venue" + "thumbnail award" + "thumbnail links"; + grid-template-columns: 300px 1fr; + gap: 10px; + + @media (max-width: 600px) { + grid-template-areas: + "title" + "authors" + "venue" + "award" + "links" + "thumbnail"; + grid-template-columns: 1fr; + } } } @@ -40,22 +47,22 @@ article.publication-card { .publication-title { grid-area: title; margin: 0; + padding: 0; + border-bottom: none; + font-size: 1.05rem; + line-height: 1.35; >a { - color: $primary-color; text-decoration: none; - - &:hover { - color: darken($primary-color, 10%); - text-decoration: none; - } } } .publication-authors { grid-area: authors; - font-style: italic; - margin: 0; + margin: 0.5rem 0 0; + font-size: 0.8rem; + line-height: 1.6; + color: #555; white-space: normal; overflow: visible; word-wrap: break-word; @@ -63,8 +70,10 @@ article.publication-card { .publication-venue { grid-area: venue; - font-weight: bold; margin: 0; + font-size: 0.75rem; + font-weight: 600; + color: #333; } .publication-links { @@ -141,4 +150,47 @@ article.publication-card { margin-top: 2em; padding-bottom: 0.5em; border-bottom: 1px solid #f2f3f3; -} \ No newline at end of file +} +/* Award note after the venue (set `award:` and optionally `award_url:` in a paper's front matter) */ +.publication-award { + font-style: italic; + font-weight: normal; + + a { + color: inherit; + text-decoration: none; + + &:hover { + text-decoration: underline; + } + } +} + +/* Back link on individual paper pages (_layouts/publication.html); a script moves it above the title. */ +.back-to-publications { + display: inline-flex; + align-items: center; + gap: 0.4rem; + margin-bottom: 0.75rem; + font-size: 0.8rem; + font-weight: 600; + color: #555; + text-decoration: none !important; + + &:hover, + &:focus-visible { + color: $primary-color; + } +} + +/* Paper / Code / Page links: same plain style as the research-page cards */ +article.publication-card .publication-links { + font-size: 0.75rem; +} + +/* Publications page: comfortable reading width, like the research pages */ +.publications-page #main>.splash { + max-width: 960px; + margin: 0 auto; + padding: 0 1rem; +} diff --git a/_sass/custom/display-tech-transfer.scss b/_sass/custom/display-tech-transfer.scss index 10622c45d..7cda94ffe 100644 --- a/_sass/custom/display-tech-transfer.scss +++ b/_sass/custom/display-tech-transfer.scss @@ -1,32 +1,192 @@ +/* Tools & resources entries (Tools & Resources page and research direction pages) */ + .tech-transfer__wrapper { + display: flex; + flex-direction: column; + gap: 1rem; margin-bottom: 2em; .tech-transfer-item { - margin-bottom: 3em; - padding: 1em; + padding: 1.5rem; background: #fff; - border-radius: 4px; - transition: all 0.3s cubic-bezier(.25, .8, .25, 1); + border: 1px solid rgba(0, 0, 0, 0.07); + border-radius: 12px; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04); h2 { - margin-top: 0; + margin: 0 0 1rem; + padding: 0; + border-bottom: none; + font-size: 1.2rem; color: $primary-color; + + a { + text-decoration: none; + } + } + + p { + margin: 0 0 0.75rem; + font-size: 0.9rem; + line-height: 1.65; + + &:last-child { + margin-bottom: 0; + } } img { - max-width: 200px; + // Never upscale small logos; cap large images at 200px. + width: auto; + max-width: min(200px, 40%); float: left; - margin: 0 1em 1em 0; + margin: 0.2rem 1.5rem 0.5rem 0; + border-radius: 6px; } .clearfix { - clear: both; - &::after { content: ""; clear: both; display: table; } } + + @media (max-width: 600px) { + img { + float: none; + display: block; + width: 100%; + max-width: 100%; + margin: 0 0 1rem; + } + } + } +} + +/* Link buttons and pip install chip under each entry */ + +.tool-links { + clear: both; + display: flex; + flex-wrap: wrap; + gap: 0.5rem; + margin-top: 1.25rem; +} + +.tool-link, +.tool-pip { + display: inline-flex; + align-items: center; + gap: 0.5rem; + padding: 0.45rem 0.9rem; + border: 1px solid rgba(155, 58, 43, 0.25); + border-radius: 999px; + background: #fff; + color: #9b3a2b; + font-size: 0.75rem; + font-weight: 600; + line-height: 1.2; + text-decoration: none !important; + transition: background-color 0.15s ease, border-color 0.15s ease, color 0.15s ease; + + &:hover, + &:focus-visible { + background: #f6efed; + border-color: rgba(155, 58, 43, 0.45); + color: #9b3a2b; + } +} + +.tool-link--primary { + background: #9b3a2b; + border-color: #9b3a2b; + color: #fff; + + &:hover, + &:focus-visible { + background: #7f2f23; + border-color: #7f2f23; + color: #fff; + } +} + +.tool-pip { + border-color: #e4e4e4; + background: #f7f7f8; + color: #333; + cursor: pointer; + font-family: inherit; + + code { + padding: 0; + background: none; + font-size: 0.75rem; + color: inherit; + } + + .tool-pip__icon { + color: #999; + } + + &:hover, + &:focus-visible { + background: #efeff1; + border-color: #d6d6d6; + color: #333; + } + + &.is-copied { + border-color: #6aa36f; + + .tool-pip__icon::before { + content: "\f00c"; // check mark + font-weight: 900; + color: #3d8b44; + } + } +} + +/* Sort dropdown above the Tools & Resources list */ + +.tool-sort { + display: flex; + justify-content: flex-end; + margin: 0 0 0.5rem; + + &__control { + position: relative; + display: inline-flex; + align-items: center; + color: #555; + font-size: 0.8rem; + + .fa-chevron-down { + position: absolute; + right: 0.6rem; + font-size: 0.65rem; + pointer-events: none; + } + } + + &__select { + appearance: none; + -webkit-appearance: none; + margin: 0; + padding: 0.35rem 1.8rem 0.35rem 0.6rem; + border: none; + border-radius: 6px; + background: transparent; + color: inherit; + font: inherit; + font-weight: 600; + cursor: pointer; + box-shadow: none; + + &:hover, + &:focus-visible { + background: #f3f3f3; + outline: none; + } } -} \ No newline at end of file +} diff --git a/_sass/custom/funders-carousel.scss b/_sass/custom/funders-carousel.scss new file mode 100644 index 000000000..524f2f476 --- /dev/null +++ b/_sass/custom/funders-carousel.scss @@ -0,0 +1,94 @@ +/* + Funders carousel (homepage) + ========================================================================== */ + +@keyframes fundersScroll { + from { + transform: translateX(0); + } + + to { + // The track holds the logo list twice, so shifting by half loops seamlessly. + transform: translateX(-50%); + } +} + +.funders-carousel { + margin: 6rem 0 3rem; + text-align: center; + + &__title { + margin-bottom: 1.5rem; + padding-bottom: 0; + border-bottom: none; + font-size: $type-size-4; + color: $text-color; + } + + &__viewport { + max-width: 760px; + margin: 0 auto; + overflow: hidden; + mask-image: linear-gradient(to right, transparent, #000 10%, #000 90%, transparent); + -webkit-mask-image: linear-gradient(to right, transparent, #000 10%, #000 90%, transparent); + } + + &__track { + display: flex; + width: max-content; + margin: 0; + padding: 0; + list-style: none; + animation: fundersScroll 40s linear infinite; + } + + &__item { + flex: 0 0 auto; + display: flex; + align-items: center; + justify-content: center; + height: 70px; + margin: 0 2.5rem; + + img { + max-height: 100%; + max-width: 160px; + width: auto; + object-fit: contain; + // filter/opacity are set per frame by the script in funders-carousel.html + will-change: filter, opacity; + } + } + + @media (max-width: 768px) { + &__item { + height: 50px; + margin: 0 1.5rem; + + img { + max-width: 110px; + } + } + } + + // No auto-scroll for reduced motion: show one copy and let users swipe/scroll. + @media (prefers-reduced-motion: reduce) { + &__viewport { + overflow-x: auto; + } + + &__track { + animation: none; + } + + &__item[aria-hidden="true"] { + display: none; + } + } +} + +// Override the theme's underline on content h2s +.page__content .funders-carousel__title { + padding-bottom: 0; + border-bottom: none; +} diff --git a/_sass/custom/header-footer.scss b/_sass/custom/header-footer.scss index 0909ac243..b1a896570 100644 --- a/_sass/custom/header-footer.scss +++ b/_sass/custom/header-footer.scss @@ -1,19 +1,63 @@ +$navbar-bg-color: #faf9f6; +$navbar-text-color: #9b3a2b; + a.site-title { + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; + @media (min-width: 601px) { - font-size: xx-large; + font-size: 1.35rem; } @media (max-width: 600px) { - font-size: large; + font-size: 0.95rem; } - color: $primary-color; + color: $navbar-text-color; + line-height: 1.1; &:hover { - color: mix($background-color, $primary-color, 25%); + color: mix(#000, $navbar-text-color, 20%); } } +.masthead, +.greedy-nav { + background-color: $navbar-bg-color; +} + +.masthead { + padding-top: 0.15rem; + padding-bottom: 0.15rem; +} + +.masthead .masthead__inner-wrap { + max-width: 100%; + width: 100%; + padding-top: 0.15rem; + padding-bottom: 0.15rem; + padding-left: 1rem; + padding-right: 1rem; +} + +.greedy-nav { + min-height: 2.6rem; + line-height: 2.6rem; +} + +.greedy-nav .visible-links a { + padding-top: 0; + padding-bottom: 0; + line-height: 2.6rem; +} + +.greedy-nav a { + color: $navbar-text-color; +} + +.greedy-nav a:hover { + color: mix(#000, $navbar-text-color, 20%); +} + .theme-toggle { @media (min-width: 601px) { margin: 0px; @@ -21,22 +65,72 @@ a.site-title { } +/* Dropdown menus + ========================================================================== */ + .submenu { list-style: none; } -.masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li:hover>.submenu { +.masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li:hover>.submenu, +.masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li:focus-within>.submenu { display: block; } .masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li>.submenu { position: absolute; - background: $background-color; - min-width: 200px; - z-index: 1000; display: none; + z-index: 1000; + min-width: 250px; + margin: 0 0 0 0.25rem; + padding: 0.5rem; + background: #fff; + border: 1px solid rgba(0, 0, 0, 0.06); + border-radius: 10px; + box-shadow: 0 10px 30px rgba(0, 0, 0, 0.08), 0 2px 6px rgba(0, 0, 0, 0.04); + line-height: 1.4; + + >li>a { + margin: 0; + padding: 0.5rem 0.75rem; + border-radius: 6px; + font-size: 0.9rem; + line-height: 1.4; + color: #2b2b2b; + white-space: nowrap; + transition: background-color 0.15s ease, color 0.15s ease; + + // No navbar underline animation inside the dropdown + &::before { + display: none; + } + + &:hover, + &:focus-visible { + background: #f6efed; + color: $navbar-text-color; + } + } +} + +// Section label and separator (`heading:` / `divider:` entries in _data/navigation.yml) +.submenu__heading { + padding: 0.5rem 0.75rem 0.35rem; + font-size: 0.68rem; + font-weight: 600; + letter-spacing: 0.12em; + text-transform: uppercase; + line-height: 1.4; + color: $gray; + cursor: default; +} + +.submenu__divider { + height: 1px; + margin: 0.4rem 0.75rem; + background: #ececec; } -.masthead>.masthead__inner-wrap>.masthead__menu>.greedy-nav>.visible-links>li>.submenu>li>a { - padding: 10px 20px; +.greedy-nav .hidden-links .submenu__divider { + border-bottom: none; } \ No newline at end of file diff --git a/_sass/custom/particles.scss b/_sass/custom/particles.scss index b4c606dcb..c2cc54fca 100644 --- a/_sass/custom/particles.scss +++ b/_sass/custom/particles.scss @@ -12,9 +12,11 @@ .page__hero--overlay { position: relative; - height: 300px; + // Grow with the content instead of a fixed height, so titles never crowd the navbar. + min-height: 300px; + background-color: #fff; overflow: hidden; - padding: 0 1em; + padding: 4rem 1em 3rem; display: flex; align-items: center; justify-content: center; @@ -53,7 +55,7 @@ margin: 0 auto; @media (min-width: 80em) { - padding: 3em; + padding: 0 3em; } @media (max-width: 768px) { @@ -65,12 +67,13 @@ .page_overlay_image { height: 100%; - min-height: 300px; + min-height: 220px; } .page__title-container { animation: fadeInUp 0.6s ease forwards; grid-column: 2; + position: relative; .title-line { width: 100px; @@ -91,6 +94,7 @@ .page__title { font-size: $type-size-1; + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; color: #000; margin: 0; line-height: 1.2; @@ -102,4 +106,70 @@ color: #000; margin-top: 1em; } +} + +.page__hero--particles { + // Research direction pages: icon inline before the title, text below, + // aligned with the 960px content column (see research-page.scss). + .wrapper--research { + display: block; + max-width: 960px; + padding: 0 1rem; + } + + .page__title-logo { + display: inline-block; + height: 1em; + width: auto; + margin-right: 0.35em; + vertical-align: -0.12em; + } + + .wrapper, + .page_overlay_image, + .page__title-container { + pointer-events: none; + } + + .page__title-container { + &::before { + content: ""; + position: absolute; + top: 50%; + left: 50%; + width: 118%; + height: 145%; + transform: translate(-50%, -50%); + background: radial-gradient(ellipse at center, + rgba(255, 255, 255, 1) 0%, + rgba(255, 255, 255, 0.97) 25%, + rgba(255, 255, 255, 0.8) 52%, + rgba(255, 255, 255, 0.35) 72%, + rgba(255, 255, 255, 0.1) 82%, + rgba(255, 255, 255, 0) 90%); + filter: blur(2px); + z-index: -1; + pointer-events: none; + } + } + + .page__title, + .page__lead, + .title-line { + position: relative; + z-index: 2; + } + + .page__title, + .page__lead { + color: #000 !important; + opacity: 1; + } +} + + + +h1.page__title, +.page__content>h1:first-of-type { + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; } \ No newline at end of file diff --git a/_sass/custom/people-card.scss b/_sass/custom/people-card.scss index 009a26d22..af20bd1ff 100644 --- a/_sass/custom/people-card.scss +++ b/_sass/custom/people-card.scss @@ -244,4 +244,9 @@ div.author-avatar { margin-right: .2em; font-size: x-large; color: $primary-color -} \ No newline at end of file +} +/* People page: extra room before the Past Members section */ +.page__content hr + #past-members, +#past-members { + margin-top: 4.5rem; +} diff --git a/_sass/custom/research-directions.scss b/_sass/custom/research-directions.scss index 1aaa71f75..6e9459ffb 100644 --- a/_sass/custom/research-directions.scss +++ b/_sass/custom/research-directions.scss @@ -2,6 +2,51 @@ Research Directions ========================================================================== */ +@keyframes logoFloat { + 0%, + 100% { + transform: translateY(0); + } + + 50% { + transform: translateY(-5px); + } +} + +.research-directions-section { + position: relative; +} + +.research-directions-section--with-particles { + margin-top: 2.5rem; + width: 100vw; + margin-left: calc(50% - 50vw); + margin-right: calc(50% - 50vw); + + #particles-js { + position: absolute; + top: 0; + left: 0; + width: 100%; + height: 100%; + z-index: 0; + background-color: #faf9f6; + } + + .research-directions__wrapper { + position: relative; + z-index: 1; + max-width: 1200px; + margin-left: auto; + margin-right: auto; + padding-top: 1.5em; + padding-left: 1em; + padding-right: 1em; + border-radius: 10px; + pointer-events: none; + } +} + .research-directions__wrapper { @include clearfix(); padding-bottom: 1em; @@ -42,6 +87,151 @@ } } +.research-directions-item--home { + box-shadow: none; + border-radius: 0; + min-height: 0; + padding-bottom: 1.1rem; + pointer-events: none; + + &:hover { + transform: none; + } + + a { + position: relative; + display: flex; + flex-direction: column; + align-items: center; + justify-content: flex-start; + text-align: center; + height: auto; + overflow: visible; + background: transparent; + padding: 1.1rem 0.5rem 0.35rem; + isolation: isolate; + pointer-events: auto; + } + + .direction-logo-wrap { + position: relative; + z-index: 2; + display: inline-flex; + align-items: center; + justify-content: center; + margin-bottom: 0.9rem; + animation: logoFloat 4.4s ease-in-out infinite; + + &::before { + content: ""; + position: absolute; + top: 50%; + left: 50%; + width: 150px; + height: 150px; + transform: translate(-50%, -50%); + background: radial-gradient(ellipse at center, + rgba(255, 255, 255, 1) 0%, + rgba(255, 255, 255, 0.98) 20%, + rgba(255, 255, 255, 0.85) 42%, + rgba(255, 255, 255, 0.45) 62%, + rgba(255, 255, 255, 0.14) 76%, + rgba(255, 255, 255, 0) 86%); + filter: blur(2px); + z-index: 0; + pointer-events: none; + } + } + + @media (prefers-reduced-motion: reduce) { + .direction-logo-wrap { + animation: none; + } + } + + .direction-logo { + position: relative; + z-index: 3; + width: 95px; + height: 95px; + object-fit: contain; + transition: transform 220ms ease; + box-shadow: none !important; + filter: none !important; + } + + a:hover .direction-logo, + a:focus .direction-logo, + a:focus-visible .direction-logo { + transform: scale(1.08); + } + + a:hover .direction-logo, + a:focus .direction-logo, + a:active .direction-logo { + box-shadow: none !important; + filter: none !important; + } + + .direction-content { + position: relative; + z-index: 1; + max-width: 320px; + margin-left: auto; + margin-right: auto; + padding: 0.35rem 0.45rem 0.5rem; + transition: transform 220ms ease; + transform-origin: center top; + + &::before { + content: ""; + position: absolute; + top: 50%; + left: 50%; + width: 140%; + height: 150%; + transform: translate(-50%, -50%); + background: radial-gradient(ellipse at center, + rgba(255, 255, 255, 1) 0%, + rgba(255, 255, 255, 0.98) 26%, + rgba(255, 255, 255, 0.82) 50%, + rgba(255, 255, 255, 0.34) 70%, + rgba(255, 255, 255, 0.1) 78%, + rgba(255, 255, 255, 0) 86%); + filter: blur(2px); + z-index: -1; + pointer-events: none; + } + } + + a:hover .direction-content, + a:focus .direction-content, + a:focus-visible .direction-content { + transform: scale(1.04); + } + + .direction-title { + justify-content: center; + border-bottom: 0; + padding-bottom: 0.2rem; + + h2 { + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; + font-size: 1.35rem; + line-height: 1.08; + } + } + + .direction-excerpt { + margin-top: 0; + font-size: 0.98rem; + + p { + line-height: 1.3; + } + } +} + .direction-image { position: relative; min-height: 250px; diff --git a/_sass/custom/research-links.scss b/_sass/custom/research-links.scss new file mode 100644 index 000000000..90992c52a --- /dev/null +++ b/_sass/custom/research-links.scss @@ -0,0 +1,88 @@ +/* + Research links (homepage cards below the research directions) + ========================================================================== */ + +html { + scroll-behavior: smooth; +} + +#research { + scroll-margin-top: 1rem; +} + +@media (prefers-reduced-motion: reduce) { + html { + scroll-behavior: auto; + } +} + +.research-links { + display: grid; + grid-template-columns: repeat(3, 1fr); + gap: 1.25rem; + max-width: 1100px; + margin: 6rem auto; + + @media (max-width: 900px) { + grid-template-columns: 1fr; + max-width: 560px; + } + + &__card, + &__card:hover, + &__card:focus-visible, + &__card * { + text-decoration: none !important; + } + + &__card { + display: flex; + align-items: center; + gap: 1rem; + padding: 1.25rem 1.5rem; + background: #fff; + border: 1px solid rgba(0, 0, 0, 0.07); + border-radius: 14px; + box-shadow: 0 1px 2px rgba(0, 0, 0, 0.04); + color: inherit; + transition: transform 0.2s ease, box-shadow 0.2s ease, border-color 0.2s ease; + + &:hover, + &:focus-visible { + transform: translateY(-3px); + border-color: rgba(155, 58, 43, 0.25); + box-shadow: 0 12px 28px rgba(0, 0, 0, 0.08); + color: inherit; + } + } + + &__icon { + flex: 0 0 auto; + display: flex; + align-items: center; + justify-content: center; + width: 2.5rem; + height: 2.5rem; + border-radius: 10px; + background: #f6efed; + color: #9b3a2b; + font-size: 1rem; + } + + &__title { + font-weight: 700; + font-size: 1rem; + line-height: 1.3; + color: #1f1f1f; + } + + @media (prefers-reduced-motion: reduce) { + &__card { + transition: none; + } + + &__card:hover { + transform: none; + } + } +} diff --git a/_sass/custom/research-page.scss b/_sass/custom/research-page.scss new file mode 100644 index 000000000..a22752860 --- /dev/null +++ b/_sass/custom/research-page.scss @@ -0,0 +1,31 @@ +/* + Research direction pages (body.research-page, set in _config.yml defaults) + ========================================================================== */ + +.research-page { + .splash>.page__content { + max-width: 960px; + margin: 0 auto; + padding: 0 1rem 2rem; + + >p { + line-height: 1.75; + } + + >h1 { + margin: 3.5rem 0 1.25rem; + padding-bottom: 0.6rem; + border-bottom: 1px solid #ececec; + font-family: 'DM Sans', -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif; + font-size: 1.5rem; + } + + >h1:first-of-type { + margin-top: 2.5rem; + } + } + + .logo-grid { + padding: 1.25rem 0; + } +} diff --git a/_sass/minimal-mistakes.scss b/_sass/minimal-mistakes.scss index 969675c81..c92eb03f5 100644 --- a/_sass/minimal-mistakes.scss +++ b/_sass/minimal-mistakes.scss @@ -52,7 +52,10 @@ @import "custom/research-directions.scss"; @import "custom/sub_research_directions.scss"; @import "custom/team-grid.scss"; -@import "custom/logo-grid.scss"; +@import "custom/logo-grid.scss"; +@import "custom/funders-carousel.scss"; +@import "custom/research-links.scss"; +@import "custom/research-page.scss"; @import "custom/display-highlighted-publications.scss"; @import "custom/breadcrumbs.scss"; @import "custom/particles.scss"; \ No newline at end of file diff --git a/_tech_transfer/credibench.md b/_tech_transfer/credibench.md new file mode 100644 index 000000000..63710ebb8 --- /dev/null +++ b/_tech_transfer/credibench.md @@ -0,0 +1,23 @@ +--- +category: information-integrity +paper_date: 2026-08-09 # date of the most recent paper; entries are listed newest first +pip: credigraph +links: + - label: Domain query API docs + url: https://credinet.readthedocs.io/en/latest/ + type: docs + - label: Dataset + url: https://huggingface.co/datasets/ComplexDataLab/CrediBench + type: dataset + - label: Paper (KDD 2026) + url: https://dl.acm.org/doi/10.1145/3770855.3817476 + type: paper +--- + +## [CrediBench](https://dl.acm.org/doi/10.1145/3770855.3817476) + +
+[{: .align-left}](https://dl.acm.org/doi/10.1145/3770855.3817476)
+
+The CrediBench dataset is a temporal benchmark that consists of monthly webgraph snapshots built from Common Crawl, where each node is a web domain with its textual page content. Each monthly snapshot contains dozens of millions of domains and over a billion edges (hyperlink relations between them). The optimized graph construction pipeline is open-sourced. Trained on this data, we make available the reliability predictions by our graph- and text-based model. We also released a new large-scale set of labelled data spanning web categories.
+
diff --git a/_tech_transfer/infrared.md b/_tech_transfer/infrared.md
index 034ace0c0..9cacd6e13 100644
--- a/_tech_transfer/infrared.md
+++ b/_tech_transfer/infrared.md
@@ -1,6 +1,5 @@
---
category: online-crime
-order: 2
---
## Infrared
diff --git a/_tech_transfer/misinfo-datasets.md b/_tech_transfer/misinfo-datasets.md
index d8bf12958..ef92dea0c 100644
--- a/_tech_transfer/misinfo-datasets.md
+++ b/_tech_transfer/misinfo-datasets.md
@@ -1,6 +1,16 @@
---
category: information-integrity
-order: 2
+paper_date: 2025-08-03 # date of the most recent paper; entries are listed newest first
+links:
+ - label: Explore the datasets
+ url: https://misinfo-datasets.complexdatalab.com/
+ type: website
+ - label: Code
+ url: https://github.com/ComplexData-MILA/misinfo-datasets
+ type: code
+ - label: Paper (KDD 2025)
+ url: https://dl.acm.org/doi/10.1145/3711896.3737437
+ type: paper
---
## [Misinfo Datasets](https://misinfo-datasets.complexdatalab.com/)
diff --git a/_tech_transfer/openfake.md b/_tech_transfer/openfake.md
new file mode 100644
index 000000000..edc5c1eac
--- /dev/null
+++ b/_tech_transfer/openfake.md
@@ -0,0 +1,22 @@
+---
+category: poli-sci
+paper_date: 2025-09-11 # date of the most recent paper; entries are listed newest first
+links:
+ - label: Deepfake detector
+ url: https://vicliv.ca/deepfake-detector/
+ type: demo
+ - label: Dataset
+ url: https://huggingface.co/datasets/ComplexDataLab/OpenFake
+ type: dataset
+ - label: Paper
+ url: https://arxiv.org/abs/2509.09495
+ type: paper
+---
+
+## [OpenFake](https://vicliv.ca/deepfake-detector/)
+
+
+[{: .align-left}](https://vicliv.ca/deepfake-detector/)
+
+OpenFake is an open dataset of real and AI-generated images from a wide range of open-source and proprietary generators, designed for training and evaluating AI-image detectors. We use it to develop detectors that recognize realistic AI-generated images, including those shared on social media.
+
diff --git a/_tech_transfer/social-simulation.md b/_tech_transfer/social-simulation.md
index 49191089b..5cb1ffcef 100644
--- a/_tech_transfer/social-simulation.md
+++ b/_tech_transfer/social-simulation.md
@@ -1,11 +1,18 @@
---
category: poli-sci
-order: 3
+paper_date: 2024-10-17 # date of the most recent paper; entries are listed newest first
+links:
+ - label: Code
+ url: https://github.com/sandbox-social/silisocs
+ type: code
+ - label: Paper
+ url: https://arxiv.org/abs/2410.13915
+ type: paper
---
-## [Social Simulation](https://github.com/social-sandbox/mastodon-sim)
+## [Social Simulation](https://github.com/sandbox-social/silisocs)
-[{: .align-left}](https://github.com/social-sandbox/mastodon-sim)
+[{: .align-left}](https://github.com/sandbox-social/silisocs)
Multi-agent simulator based on Deepmind's Concordia, a software library for LLM-based multi-agent simulations of real world human experience.
diff --git a/_tech_transfer/temporal-graph-benchmark.md b/_tech_transfer/temporal-graph-benchmark.md
index f19252879..d643d5c69 100644
--- a/_tech_transfer/temporal-graph-benchmark.md
+++ b/_tech_transfer/temporal-graph-benchmark.md
@@ -1,6 +1,20 @@
---
category: TGL
-order: 1
+paper_date: 2024-06-14 # date of the most recent paper; entries are listed newest first
+pip: py-tgb
+links:
+ - label: Website & leaderboards
+ url: https://tgb.complexdatalab.com/
+ type: website
+ - label: Code
+ url: https://github.com/shenyangHuang/TGB
+ type: code
+ - label: TGB paper (NeurIPS 2023)
+ url: https://arxiv.org/abs/2307.01026
+ type: paper
+ - label: TGB 2.0 paper (NeurIPS 2024)
+ url: https://arxiv.org/abs/2406.09639
+ type: paper
---
## [Temporal Graph Benchmark](https://tgb.complexdatalab.com/)
diff --git a/_tech_transfer/tgm.md b/_tech_transfer/tgm.md
new file mode 100644
index 000000000..f675f40bd
--- /dev/null
+++ b/_tech_transfer/tgm.md
@@ -0,0 +1,23 @@
+---
+category: TGL
+paper_date: 2026-04-23 # date of the most recent paper; entries are listed newest first
+pip: tgm-lib
+links:
+ - label: Documentation
+ url: https://tgm.readthedocs.io
+ type: docs
+ - label: Code
+ url: https://github.com/tgm-team/tgm
+ type: code
+ - label: Paper (ICLR 2026)
+ url: https://arxiv.org/abs/2510.07586
+ type: paper
+---
+
+## [TGM: Temporal Graph Modelling](https://github.com/tgm-team/tgm)
+
+
+[{: .align-left}](https://github.com/tgm-team/tgm)
+
+TGM is a research-oriented Python library for machine learning on temporal graphs, and the first to unify continuous- and discrete-time dynamic graph methods. It offers first-class support for dynamic node features, time-granularity conversions, and link-, node- and graph-level tasks, and runs on average 7.8x faster than the widely used DyGLib.
+
diff --git a/_tech_transfer/veracity-ai.md b/_tech_transfer/veracity-ai.md
index e0efcc1c8..76f6d4897 100644
--- a/_tech_transfer/veracity-ai.md
+++ b/_tech_transfer/veracity-ai.md
@@ -1,6 +1,16 @@
---
category: poli-sci
-order: 1
+paper_date: 2025-06-18 # date of the most recent paper; entries are listed newest first
+links:
+ - label: Try Veracity
+ url: https://www.veri-fact.ai/
+ type: demo
+ - label: Chrome Extension
+ url: https://chromewebstore.google.com/detail/veracity/hoapdcnhiffbhkdnkpamhdapcljidbpp
+ type: extension
+ - label: Paper
+ url: https://arxiv.org/abs/2506.15794
+ type: paper
---
## [Veracity AI](https://www.veri-fact.ai/)
diff --git a/assets/images/bio/emma.kondrup.avatar.webp b/assets/images/bio/emma.kondrup.avatar.webp
new file mode 100644
index 000000000..c3c0d6df6
Binary files /dev/null and b/assets/images/bio/emma.kondrup.avatar.webp differ
diff --git a/assets/images/home/AI4.png b/assets/images/home/AI4.png
new file mode 100644
index 000000000..1cafcd87b
Binary files /dev/null and b/assets/images/home/AI4.png differ
diff --git a/assets/images/home/TGL.png b/assets/images/home/TGL.png
new file mode 100644
index 000000000..08f91b52e
Binary files /dev/null and b/assets/images/home/TGL.png differ
diff --git a/assets/images/home/group-photo-2.webp b/assets/images/home/group-photo-2.webp
index deb3a8ff5..70e503c49 100644
Binary files a/assets/images/home/group-photo-2.webp and b/assets/images/home/group-photo-2.webp differ
diff --git a/assets/images/home/huggingface.webp b/assets/images/home/huggingface.webp
new file mode 100644
index 000000000..79b901645
Binary files /dev/null and b/assets/images/home/huggingface.webp differ
diff --git a/assets/images/home/infrared.png b/assets/images/home/infrared.png
new file mode 100644
index 000000000..ecec859bf
Binary files /dev/null and b/assets/images/home/infrared.png differ
diff --git a/assets/images/home/misinfo_gov_outreach.png b/assets/images/home/misinfo_gov_outreach.png
deleted file mode 100644
index b80730f33..000000000
Binary files a/assets/images/home/misinfo_gov_outreach.png and /dev/null differ
diff --git a/assets/images/home/misinfo_gov_outreach.webp b/assets/images/home/misinfo_gov_outreach.webp
index 21f0fe887..1b922dfcb 100644
Binary files a/assets/images/home/misinfo_gov_outreach.webp and b/assets/images/home/misinfo_gov_outreach.webp differ
diff --git a/assets/images/home/toxicity.png b/assets/images/home/toxicity.png
new file mode 100644
index 000000000..0e10de483
Binary files /dev/null and b/assets/images/home/toxicity.png differ
diff --git a/assets/images/logo/AISI.png b/assets/images/logo/AISI.png
new file mode 100644
index 000000000..6e5669d99
Binary files /dev/null and b/assets/images/logo/AISI.png differ
diff --git a/assets/images/logo/CANADA.png b/assets/images/logo/CANADA.png
index c821c6aa4..dfed1e1b5 100644
Binary files a/assets/images/logo/CANADA.png and b/assets/images/logo/CANADA.png differ
diff --git a/assets/images/logo/FLI.png b/assets/images/logo/FLI.png
new file mode 100644
index 000000000..01e084077
Binary files /dev/null and b/assets/images/logo/FLI.png differ
diff --git a/assets/images/logo/MILA.png b/assets/images/logo/MILA.png
index 3c5453a1a..f934410cc 100644
Binary files a/assets/images/logo/MILA.png and b/assets/images/logo/MILA.png differ
diff --git a/assets/images/logo/SSHRC.png b/assets/images/logo/SSHRC.png
index c9de5b524..6b6ba6b6b 100644
Binary files a/assets/images/logo/SSHRC.png and b/assets/images/logo/SSHRC.png differ
diff --git a/assets/images/logo/mitacs.png b/assets/images/logo/mitacs.png
index 60e5512cc..9eb68e190 100644
Binary files a/assets/images/logo/mitacs.png and b/assets/images/logo/mitacs.png differ
diff --git a/assets/images/logo/ubisoft-la-forge.png b/assets/images/logo/ubisoft-la-forge.png
index 2db58eb31..baebc822a 100644
Binary files a/assets/images/logo/ubisoft-la-forge.png and b/assets/images/logo/ubisoft-la-forge.png differ
diff --git a/assets/images/tech_transfer/credibench.webp b/assets/images/tech_transfer/credibench.webp
new file mode 100644
index 000000000..a80027a0d
Binary files /dev/null and b/assets/images/tech_transfer/credibench.webp differ
diff --git a/assets/images/tech_transfer/openfake.webp b/assets/images/tech_transfer/openfake.webp
new file mode 100644
index 000000000..96c6e4de3
Binary files /dev/null and b/assets/images/tech_transfer/openfake.webp differ
diff --git a/assets/images/tech_transfer/tgm.webp b/assets/images/tech_transfer/tgm.webp
new file mode 100644
index 000000000..c1da1bd83
Binary files /dev/null and b/assets/images/tech_transfer/tgm.webp differ
diff --git a/assets/js/particles-config.js b/assets/js/particles-config.js
index 54719b156..42d0fc9ed 100644
--- a/assets/js/particles-config.js
+++ b/assets/js/particles-config.js
@@ -66,4 +66,33 @@ document.addEventListener('DOMContentLoaded', function() {
},
"retina_detect": true
});
+
+ // On the homepage research section, foreground content can block direct canvas
+ // mouse events. Mirror pointer coordinates into particles.js so hover effects
+ // remain interactive across the full section area.
+ var particlesEl = document.getElementById('particles-js');
+ var particlesSection = particlesEl && particlesEl.closest('.research-directions-section--with-particles');
+
+ if (particlesEl && particlesSection && window.pJSDom && window.pJSDom[0] && window.pJSDom[0].pJS) {
+ var pJS = window.pJSDom[0].pJS;
+
+ particlesSection.addEventListener('mousemove', function(event) {
+ var canvas = pJS.canvas && pJS.canvas.el;
+ if (!canvas) {
+ return;
+ }
+
+ var rect = canvas.getBoundingClientRect();
+ var pxratio = (pJS.canvas && pJS.canvas.pxratio) ? pJS.canvas.pxratio : 1;
+ pJS.interactivity.mouse.pos_x = (event.clientX - rect.left) * pxratio;
+ pJS.interactivity.mouse.pos_y = (event.clientY - rect.top) * pxratio;
+ pJS.interactivity.status = 'mousemove';
+ });
+
+ particlesSection.addEventListener('mouseleave', function() {
+ pJS.interactivity.mouse.pos_x = null;
+ pJS.interactivity.mouse.pos_y = null;
+ pJS.interactivity.status = 'mouseleave';
+ });
+ }
});
\ No newline at end of file
diff --git a/src/python/add_update_member.py b/src/python/add_update_member.py
index 01275c655..af033fbec 100644
--- a/src/python/add_update_member.py
+++ b/src/python/add_update_member.py
@@ -6,6 +6,14 @@
from . import save_url_image, parse_issue_body, remove_keys, remove_items_with_values
+# What GitHub issue forms send for fields left blank: "_No response_" for text inputs,
+# "None" for dropdowns.
+BLANK_VALUES = {"", "_No response_", "None"}
+
+
+def is_filled(value) -> bool:
+ return value is not None and str(value).strip() not in BLANK_VALUES
+
def format_site_label(name: str) -> str:
"""Format social media site labels."""
@@ -23,7 +31,7 @@ def format_social_media_links(parsed: Dict) -> List[Dict]:
"""Format social media links into structured format."""
social_links = []
for key in ["website", "twitter", "github", "scholar", "linkedin"]:
- if parsed.get(key) and parsed[key] != "_No response_":
+ if is_filled(parsed.get(key)):
social_links.append({
"label": format_site_label(key),
"url": parsed[key]
@@ -32,29 +40,19 @@ def format_social_media_links(parsed: Dict) -> List[Dict]:
def process_role_data(parsed: Dict, prefix: str = "") -> Optional[Dict]:
- """Process role data from parsed content with optional prefix."""
- role_type = parsed.get(f"{prefix}type")
- if not role_type or role_type == "_No response_":
- return None
-
- role = {
- "type": role_type,
- "title": parsed.get(f"{prefix}title")
- }
+ """Process role data from parsed content with optional prefix.
- # Add optional fields if they exist
- for field in ["start_date", "end_date", "advisor"]:
+ Only fields that were filled in are included, so an update can change part of a role.
+ Returns None if no role field was filled in.
+ """
+ role = {}
+ for field in ["type", "title", "start_date", "end_date", "advisor", "affiliation"]:
value = parsed.get(f"{prefix}{field}")
- if value and value != "_No response_":
+ if is_filled(value):
role[field] = value
- # Add new fields for affiliation and research directions
- affiliation = parsed.get(f"{prefix}affiliation")
- if affiliation and affiliation != "_No response_":
- role["affiliation"] = affiliation
-
research_directions = parsed.get(f"{prefix}research_directions")
- if research_directions and research_directions != "_No response_":
+ if is_filled(research_directions):
# Handle both string and list inputs for research directions
if isinstance(research_directions, str):
# Split by commas if it's a comma-separated string
@@ -63,7 +61,7 @@ def process_role_data(parsed: Dict, prefix: str = "") -> Optional[Dict]:
directions = research_directions
role["research_directions"] = directions
- return role
+ return role or None
def format_parsed_content(parsed: Dict) -> Dict:
@@ -78,15 +76,15 @@ def format_parsed_content(parsed: Dict) -> Dict:
if current_role:
formatted["current_role"] = current_role
- # Process past role if it exists (for updates)
+ # Process past role if it exists (for updates); it needs at least a type to be identifiable
past_role = process_role_data(parsed, "past_role_")
- if past_role:
+ if past_role and "type" in past_role:
formatted["past_roles"] = past_role
# Add optional fields if they exist and aren't empty
optional_fields = ["bio", "note", "semantic_scholar_id", "avatar"]
for field in optional_fields:
- if parsed.get(field) and parsed[field] != "_No response_":
+ if is_filled(parsed.get(field)):
formatted[field] = parsed[field]
# Format social media links
@@ -106,9 +104,21 @@ def merge_profile_data(old_profile: Dict, new_profile: Dict) -> Dict:
if field in new_profile:
merged[field] = new_profile[field]
- # Update current role if provided
+ # Update current role if provided. A different type or title means a new role, which
+ # replaces the old one; otherwise only the fields filled in are changed and the rest kept.
if "current_role" in new_profile:
- merged["current_role"] = new_profile["current_role"]
+ old_role = merged.get("current_role") or {}
+ new_role = new_profile["current_role"]
+ is_new_role = any(
+ field in new_role and new_role[field] != old_role.get(field)
+ for field in ["type", "title"]
+ )
+ if is_new_role:
+ merged["current_role"] = new_role
+ else:
+ merged_role = dict(old_role)
+ merged_role.update(new_role)
+ merged["current_role"] = merged_role
# Handle past roles
if "past_roles" in new_profile:
@@ -119,7 +129,7 @@ def merge_profile_data(old_profile: Dict, new_profile: Dict) -> Dict:
role_exists = False
for role in existing_past_roles:
if (role.get("type") == new_past_role["type"] and
- role.get("title") == new_past_role["title"] and
+ role.get("title") == new_past_role.get("title") and
role.get("start_date") == new_past_role.get("start_date")):
role.update(new_past_role)
role_exists = True
@@ -160,19 +170,20 @@ def main(parsed: Dict, action: str, site_data_dir: str = "_data/", image_dir: st
with open(site_data_dir / "authors.yml") as f:
authors = yaml.load(f)
- if action == "Add member":
- # Handle new member addition
- if profile["name"] in authors:
+ name_to_username = {authors[username]["name"]: username for username in authors}
+
+ if action == "Add member" and profile["name"] not in name_to_username:
+ username = profile["name"]
+ if username in authors:
n = 2
while (k := f'{profile["name"]} {n}') in authors:
n += 1
username = k
- else:
- username = profile["name"]
authors[username] = profile
else:
- # Handle member update
- name_to_username = {authors[username]["name"]: username for username in authors}
+ # Update, or an "Add member" for someone already listed (e.g. a placeholder
+ # entry created by a maintainer): merge into the existing entry instead of
+ # creating a duplicate.
if profile["name"] not in name_to_username:
raise ValueError(f'{profile["name"]} not in authors')