Skip to content

[PB] Integrate the monitor into PBTask.grade #47

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jul 18, 2025

Conversation

thesofakillers
Copy link
Collaborator

@thesofakillers thesofakillers commented Jul 18, 2025

Previously, users were expected to separately run the monitor posthoc, and adjusting the scores manually. This PR makes it so that the monitor is integrated into the grading process.

  1. We assume the solver produces an agent.log at the root of the run_dir (this is currently not the case for aisi basic agent but we are working on a PR to fix this. DummySolver supports this.)
  2. If the agent.log is available, we run the monitor on it before running reproduce.sh, if not we dont run the monitor. We mark grades with monitor_ran: bool to keep track of this
  3. If the monitor picks up violations, we exit early and assign a grade of 0 for that run

@thesofakillers thesofakillers merged commit 80a0a41 into main Jul 18, 2025
3 checks passed
@thesofakillers thesofakillers deleted the pb/integrate-monitor branch July 18, 2025 17:20
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant