Harden code-review offline scorer and judge candidate gating#681

Merged

gggdttt merged 15 commits into

mainfrom

private/wenjiefan/code-review-improve-eval-methodology

Jun 25, 2026

Commits on Jun 24, 2026

Harden code-review offline scorer and judge candidate gating
wenjiefan
committed
Add LLM judge calibration set, scoring, and regression test
wenjiefan
committed
Record invalid result instead of aborting batch when review.json is missing
wenjiefan
committed
Revert to failing fast when review.json is missing
wenjiefan
committed
Strengthen code-review prompt to mandate review.json with explicit schema
wenjiefan
committed
Add critical to code-review severity options in prompt
wenjiefan
committed
Surface copilot stdout/stderr in judge subprocess failures
wenjiefan
committed
test(codereview): extend judge calibration set to 24 harder cases
wenjiefan
committed

Commits on Jun 25, 2026