Change the attack interface to work on the entire dataset#88
Draft
Change the attack interface to work on the entire dataset#88
Conversation
meta-cla
bot
added
the
CLA Signed
facebook-github-bot
pushed a commit
that referenced
this pull request
Feb 6, 2026
Summary: For many attack types, the attack itself should have control over the entire dataset and rollouts. So here we update the interface accordingly and introduce a rollout executor to manage that. We also wrap the simple attacks under this more new interface. **Note**: Right now, I am putting this up as a draft for architectural discussions and I have not yet checked all implementation details or tests, so the pull request may evolve. Differential Revision: D92270921 Pulled By: evtimovi
facebook-github-bot
force-pushed
the
full_dataset_attacks
branch
from
e1d6724 to
7750c9d
Compare
facebook-github-bot
pushed a commit
that referenced
this pull request
Feb 6, 2026
Summary: For many attack types, the attack itself should have control over the entire dataset and rollouts. So here we update the interface accordingly and introduce a rollout executor to manage that. We also wrap the simple attacks under this more new interface. **Note**: Right now, I am putting this up as a draft for architectural discussions and I have not yet checked all implementation details or tests, so the pull request may evolve. Differential Revision: D92270921 Pulled By: evtimovi
facebook-github-bot
force-pushed
the
full_dataset_attacks
branch
from
7750c9d to
71a8382
Compare
facebook-github-bot
pushed a commit
that referenced
this pull request
Feb 6, 2026
Summary: For many attack types, the attack itself should have control over the entire dataset and rollouts. So here we update the interface accordingly and introduce a rollout executor to manage that. We also wrap the simple attacks under this more new interface. **Note**: Right now, I am putting this up as a draft for architectural discussions and I have not yet checked all implementation details or tests, so the pull request may evolve. Differential Revision: D92270921 Pulled By: evtimovi
facebook-github-bot
force-pushed
the
full_dataset_attacks
branch
from
71a8382 to
11cf267
Compare
added 2 commits
February 6, 2026 10:43
#113) Summary: When computing pass@k metrics with k > 1, tasks that have fewer than k samples are now gracefully skipped with a warning log message instead of raising a ValueError that would terminate the entire results processing. Changes: - Replace ValueError with warning log when n_samples < k for a task - Add logging module import and logger instance - Collect skipped groups and log them with full context (dataset, agent, attack, task_id, and sample count) - Add check for empty DataFrame after filtering in aggregate_results - Update docstrings to reflect new behavior (Note instead of Raises) - Also includes refactoring: remove job_name from group_cols to allow aggregating across multiple runs of the same experiment - Add generic variant_name support alongside legacy template_short_name Differential Revision: D92393526 Pulled By: evtimovi
Summary: For many attack types, the attack itself should have control over the entire dataset and rollouts. So here we update the interface accordingly and introduce a rollout executor to manage that. We also wrap the simple attacks under this more new interface. **Note**: Right now, I am putting this up as a draft for architectural discussions and I have not yet checked all implementation details or tests, so the pull request may evolve. Differential Revision: D92270921 Pulled By: evtimovi
facebook-github-bot
force-pushed
the
full_dataset_attacks
branch
from
11cf267 to
25a9537
Compare
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.
For many attack types, the attack itself should have control over the entire dataset and rollouts. So here we update the interface accordingly and introduce a rollout executor to manage that. We also wrap the simple attacks under this more new interface.
Note: Right now, I am putting this up as a draft for architectural discussions and I have not yet checked all implementation details or tests, so the pull request may evolve.