Repair SWE test isolation and calendar overlap grading - #145
Merged
Conversation
Restore hidden-test paths from each task's immutable base commit before applying trusted patches. OpenSWE setup and patch failures now propagate without a reward, while executed test failures remain scoreable zeros. Generated SWE-ReBench tasks declare 4 GiB memory and 8 GiB storage; OpenSWE tasks declare 4 GiB storage. Calendar grading checks every overlapping event pair with half-open intervals, and generated instructions state the constraint. The v4.9 builder retains 37,952 tasks across three successor sources, removes deterministic task defects and unrecoverable calendar rows, and enforces the 20-image limit.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Restore hidden-test paths from each task's immutable base commit before applying trusted patches. OpenSWE setup and patch failures now propagate without a reward, while executed test failures remain scoreable zeros. Generated SWE-ReBench tasks declare 4 GiB memory and 8 GiB storage; OpenSWE tasks declare 4 GiB storage.
Calendar grading checks every overlapping event pair with half-open intervals, and generated instructions state the constraint. The v4.9 release retains 37,952 tasks across three successor sources, removes deterministic task defects and calendar rows without recoverable exact event names, and uses the authorized maximum of 20 images.
TaskTrove v4.9 is published at https://huggingface.co/datasets/open-thoughts/TaskTrove/tree/v4.9. The focused suite passes 90 tests; Ruff lint and format checks also pass.