Repository navigation

feat: judge reliability - calibration, uncertainty, and bias debiasing (v0.3.0) - #2

Merged

LesterALeong merged 5 commits into

feat/v0.3.0-judge-reliability

Jun 4, 2026

Merged

feat: judge reliability - calibration, uncertainty, and bias debiasing (v0.3.0)#2
LesterALeong merged 5 commits into
masterfrom
feat/v0.3.0-judge-reliability

Commits on Jun 4, 2026
- feat(judge): add JudgeDimension.score() and SelfConsistencyJudge
  
  LesterALeong
  and
  claude
  committed
  b3add53
- Add pairwise LLM-as-judge with position-bias debiasing
  
  LesterALeong
  and
  claude
  committed
  0adf8f8
- Add judge calibration module and correlation stats
  
  LesterALeong
  and
  claude
  committed
  f4befb7
- wire: export judge-reliability API, add example, README, bump 0.3.0
  LesterALeong
  committed
  80b9f7a
- docs: note self-consistency cost (N model calls per text, cache bypassed)
  LesterALeong
  committed
  fb47421