You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
review_experiment A/B output comparison in AgenticReview — runs both outputs through review_output, compares confidence scores, declares winner with delta
Specs for review_experiment covering winner selection, tie detection, delta calculation, and error handling
Changed
Add caller: { extension: 'lex-eval', operation: '...' } identity parameter to all Legion::LLM call sites: LlmJudge#evaluate_structured, LlmJudge#evaluate_regex_fallback, and AgenticReview#review_output