FFFF Add benchmark summaries and README translation policy · hmmhmmhm/aligns.md@9c8f26d · GitHub
[go: up one dir, main page]

Skip to content
  • Add benchmark summaries and README translation policy #19

    Add benchmark summaries and README translation policy

    Add benchmark summaries and README translation policy #19

    Workflow file for this run

    name: Check skill package
    on: [push, pull_request]
    permissions:
    contents: read
    jobs:
    check:
    runs-on: ubuntu-latest
    env:
    PYTHONDONTWRITEBYTECODE: '1'
    steps:
    - uses: actions/checkout@v4
    - uses: actions/setup-python@v5
    with:
    python-version: '3.12'
    - name: Check skill package
    run: python3 -m unittest discover -s tests -p 'test_*.py'
    - name: Check benchmark harness
    run: python3 -m unittest discover -s benchmarks -p 'test_*.py'
    - name: Check offline reporting
    run: python3 -m unittest discover -s benchmarks/reporting -p 'test_*.py'
    - name: Verify original review evidence without model access
    run: |
    python3 benchmarks/reporting/verify.py benchmarks/results/2026-10-15-astra-v1 \
    --judges benchmarks/results/2026-10-15-astra-v1/judgments \
    --packets benchmarks/results/2026-10-15-astra-v1/review-packets \
    --require-complete \
    --output "$RUNNER_TEMP/align-original-review-report.json"
    - name: Verify corrected primary review without model access
    run: |
    python3 benchmarks/reporting/verify.py benchmarks/results/2026-10-15-astra-v1 \
    --judges benchmarks/results/2026-10-15-astra-v1/judgments-v2 \
    --packets benchmarks/results/2026-10-15-astra-v1/review-packets-v2 \
    --require-complete \
    --output "$RUNNER_TEMP/align-evidence-report.json"
    - name: Check corrected review quotes at their declared sources
    run: |
    python3 benchmarks/reporting/check_quote_sources.py \
    benchmarks/results/2026-10-15-astra-v1/review-packets-v2 \
    benchmarks/results/2026-10-15-astra-v1/judgments-v2
    - name: Check scripted intervention harness and reporting
    run: python3 -m unittest discover -s benchmarks/effort -p 'test_*.py'
    - name: Check preference discovery harness and review
    run: python3 -m unittest discover -s benchmarks/discovery -p 'test_*.py'
    - name: Check preference review replay
    run: python3 -m unittest discover -s benchmarks/discovery/reporting -p 'test_*.py'
    - name: Verify retained intervention infrastructure failure
    run: |
    python3 benchmarks/effort/verify.py benchmarks/effort/results/2026-10-15-astra-v1-failed \
    --output "$RUNNER_TEMP/align-effort-failed-verification.json"
    - name: Verify scripted intervention results without model access
    run: |
    python3 benchmarks/effort/verify.py benchmarks/effort/results/2026-10-15-astra-v2 \
    --output "$RUNNER_TEMP/align-effort-verification.json"
    - name: Verify preference transfer results without model access
    run: |
    python3 benchmarks/discovery/verify.py benchmarks/discovery/results/2026-10-15-astra-v1 \
    --output "$RUNNER_TEMP/align-discovery-verification.json"
    - name: Verify blinded preference review without model access
    run: |
    python3 benchmarks/discovery/verify_reviews.py "$PWD/benchmarks/discovery/results/2026-10-15-astra-v1" \
    --packets "$PWD/benchmarks/discovery/results/2026-10-15-astra-v1/review-packets" \
    --reviews "$PWD/benchmarks/discovery/results/2026-10-15-astra-v1/judgments" \
    --output "$RUNNER_TEMP/align-discovery-review-verification.json"
    - name: Check preference retention protocol
    run: python3 -m unittest discover -s benchmarks/retention -p 'test_*.py'
    - name: Check retention review replay
    run: python3 -m unittest discover -s benchmarks/retention/reporting -p 'test_*.py'
    - name: Verify recovered retention evidence without model access
    run: |
    python3 benchmarks/retention/recovery/verify.py benchmarks/retention/results/2026-10-15-astra-v1 \
    --output "$RUNNER_TEMP/align-retention-verification.json"
    - name: Verify masked retention review without model access
    run: |
    python3 benchmarks/retention/reporting/verify_review.py benchmarks/retention/results/2026-10-15-astra-v1 \
    --output "$RUNNER_TEMP/align-retention-review-verification.json"
  • 0