Repository navigation
docs: corpus run 2 in agent mode — prompt A/B, recall still 0% by rule - #13
Merged
Merged
Conversation
Removing the conservative sentence from the subagent prompt moved quiet rate 100% -> 86% and noise 0.00 -> 0.14 per clean hunk; recall by rule stays 0/15 and the gate stays closed. Read by hand, 2 of 8 labelled defects were described correctly under other rules. Published as an observation, not a metric: any defect-level score is pre-registered first. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01WkPsqEJ4QYWRW5eCFBYRPS
aisona-lab
marked this pull request as ready for review
October 23, 2026 08:12
This was referenced Oct 23, 2026
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What
A second corpus run, identical to run 1 (#12) except for one change: the subagent prompt no longer says "only fail when the code shows the problem; not on speculation". This is a docs-only change, adding
docs/eval-runs/2026-10-23-agent-mode-corpus-run2.txtand a paragraph indocs/hardening-plan.md.Read by hand (not a metric, not gated on)
The reviewer described 2 of the 8 labelled fix PRs correctly, but filed both under rules other than the label:
verify=Falsebranch returns early, socertis never loaded.$also matches before a trailing newline.There are two open readings:
Any defect-level metric has to be pre-registered before the next run, not fitted to this one.
The artifact also flags one possible real defect in a "clean" hunk (
click-3876-3), not verified.Generated by Claude Code