Evals: six -llm-judge tasks graded on weighted 1-5 scales

[davidbzyk]

Adds vo-x7-llm-judge, vo-x4-llm-judge, vo-t06-llm-judge, vo-t01-llm-judge, vo-t04-llm-judge and vo-t02-llm-judge, copied byte for byte from code_review_private #20. Same seven reviewers and validator as each recall task; the grader is a judge with file tools that reads the review, the known issues and the code at the reviewed commit, and rates four 1-5 scales whose weights sum to 1: known issues 0.30, finding validity 0.25, severity and fixes 0.15, economic value 0.30. reward is their weighted mean.

The 12 existing tasks change only in steps/validator/tests/finish.py (it names the score after the rubric folder and copies a task's own judge inputs; recall tasks behave as before). README lists the new tasks; tests/test_evals.py expects the 18 tasks and checks each llm-judge grader's weights sum to 1.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[coderabbitai]

Walkthrough

Warning

Review details and warnings were omitted to fit the comment limit.

[github-actions]

Custom Code Review

1 finding on 8f7787f · 1 P2

_{codex-ci · gpt-6-luna · effort high · Codex 0.161.0}

[github-actions]

Custom Code Review

No issues found on 1f2e0f0.

What the change does

• The README now describes six LLM-judge variants of existing team review tasks and their four weighted evaluation scales.
• The eval tests include those six variants in the expected task inventory.
• A new test checks that each judge rubric has four five-point Likert criteria with the documented names and weights.
• It also checks that each task includes the known issues file and repository archive in its validator workspace.

_{codex-ci · gpt-6-luna · effort high · Codex 0.161.0}

[github-actions]

Custom Code Review

No issues found on 1f2e0f0.

_{codex-ci · gpt-6-luna · effort high · Codex 0.161.0}