Repository navigation
Per-dataset browser: size and winner columns beside the detail, and only fully measured methods - #690
Merged
Merged
Conversation
…nly fully measured methods The list beside the detail (from 1100px) shows the training rows, features and winner again, in a compact form: tighter cells, a "Rows" header, counts as 815k or 1.2M and the rank without its field size. Only the tie count is left to the row's hover and the detail. A method imputed on any split of a dataset, or without a result on one, is dropped from that dataset as a whole: its mean over the splits it ran on was ranked against means over all of them. The count of methods that could not run includes it, and no longer counts it among those that ran. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01ELdutHiUqkvynzEP7EnPsi
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The per-dataset browser's list beside the detail shows each dataset's training rows, features and winner again, in a compact form. A method imputed on any split of a dataset, or without a result on one, is now left out of that dataset as a whole, instead of being ranked on the splits it did run on.
Details
Columns: from 1100px the list sits beside the detail and hid its
widecolumns (training rows, features, tied, winner). Only the tie count stays hidden there now; it is in the row's hover and the detail. To fit the rest, the list beside the detail tightens its cell padding, shows "Rows" for "Train rows", writes counts as 815k or 1.2M (the full numbers stay in the stacked layout), drops the "/ N" after the rank, clips names and winners a little earlier, and takes 54 % of the width instead of 45 %. The strip keeps its 180px. On a 142-dataset page the list fits without scrolling from a 1440px window; at 1100 to 1280px it scrolls sideways less than before.Partial methods:
measured_resultsdropped imputed rows split by split, so a method that ran on one split of five kept a one-split mean and was ranked and tested against the methods that ran on all five (on BeyondArena core2k: TabPFN-2.6 and TabDPT on acquire_valued_shoppers_challenge). It now keeps a (dataset, method) pair only when the method has a measured result on every split of the dataset.imputed_countscounts the methods dropped that way, so the "Ran n of m methods" line no longer counts a partial method both as run and as skipped.Tests
ruff check . && ruff format --check .(ruff 0.14.0)pytest tests/tabarena/contexts tests/tabarena/plot tests/tabarena/evaluation tests/tabarena/website -q: 405 passed, with the newtest_the_list_beside_the_detail_keeps_the_size_and_winner_columnsandtest_a_method_without_a_measured_result_on_every_split_is_dropped_for_that_dataset(an imputed and a missing split).generate_all_figsrun, at 1000 to 1920px: the columns show beside the detail, the list fits from 1440px, the methods with one of five splits are gone from acquire_valued_shoppers_challenge and its field reads 39 of 47 methods; no page errors.🤖 Generated with Claude Code
https://claude.ai/code/session_01ELdutHiUqkvynzEP7EnPsi