Add llm-judges rule: judge-run methodology distilled from SPEC-B - #67
Open
yulonglin wants to merge 1 commit into
Open
Add llm-judges rule: judge-run methodology distilled from SPEC-B#67yulonglin wants to merge 1 commit into
yulonglin wants to merge 1 commit into
Conversation
Codifies the LLM-judge conventions validated by the transcript-judge e2e run: one call per sample, rationale-before-verdict with verbatim quotes, absence fields cannot be quote-grounded, state the surface, blind the judge (with the construct-naming clarification), sha-versioned prompts, append-only JSONL persistence, parse failures reported as counts. Claude-Session: https://claude.ai/code/session_01EMvJA9K5BsQFSgbuFcqCWc
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds
claude/rules/llm-judges.md— judge-run methodology distilled from the SPEC-B transcript-review build: LLM judge over regex for meaning classification, one call per sample, rationale-before-verdict, quote-grounded positives, absence-fields caveat, surface statement, blinding, sha-pinned versioned prompts, append-only resumable JSONL persistence, parse failures reported as numbers.One sentence to review closely (my addition resolving an ambiguity Codex flagged): "Naming the construct being measured is not a blinding violation — the judge must know what to look for; what it must not see is which verdict the hypothesis favours, or which experimental condition the sample came from."
https://claude.ai/code/session_01EMvJA9K5BsQFSgbuFcqCWc