Fix permissibility-split policy violation metrics - #276
Merged
AaronAspinwall123 merged 6 commits intoJul 29, 2026
Merged
Conversation
ricfog
marked this pull request as ready for review
July 24, 2026 18:44
ricfog
requested review from
AaronAspinwall123,
changliu2,
jakepresent,
minthigpen,
sooyeonni and
tangym
as code owners
July 24, 2026 18:44
Merged
The viewer stop-reason label is 'Refused before Inference' (capital I, result-view.ts:20), but three assertions in test_load_run_page_data_exposes_refusal_stop_reason_display still expected lowercase 'inference'. This mismatch is inherited from main (present at base 7a6e95b, from responsibleai#266's terminology cleanup) and fails Tier 1 unit tests. Align the test to the code label; mirrors the fix in a403385. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: ef089fbc-c563-427f-b8e1-f87370e98c76
…plit export_suite_results.py still computed not_permissible_policy_violation_rate by row-filtering on the originating permissibility label and reading the row-level policy_violation verdict - the overrefusal-contaminated approach issue responsibleai#272 calls out. Reuse compute_policy_violation_by_permissibility so the export matches the CLI/results and viewer definition: one vote per conversation over non-permissible nodes only. Keeps permissible_overrefusal_rate on the ok-filtered permissible rows. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Copilot-Session: ef089fbc-c563-427f-b8e1-f87370e98c76
AaronAspinwall123
approved these changes
Jul 28, 2026
jakepresent
approved these changes
Jul 29, 2026
ango10
pushed a commit
that referenced
this pull request
Jul 29, 2026
This was referenced Jul 30, 2026
[Bug]: Viewer defaults its headline and A/B compare axis to the coupled policy_violation metric
#291
Open
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
3 participants
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
Root cause
The CLI conditioned the overall
policy_violationverdict on the test case's originating permissibility label. The viewer grouped judgments by node permissibility but averaged individual nodes rather than computing one Boolean per conversation.Neither implementation matched the intended definition:
Impact
Overall
policy_violationremains unchanged. This does not modify judge prompts, schemas, or the evaluation pipeline.Historical results can be recalculated from their existing node judgments and taxonomy.
Validation
Fixes #272