v0.4: Add filter_supported_claims_by_relevance to normalize extraction across model sizes

v0.4: Add filter_supported_claims_by_relevance to normalize extraction across model sizes

Upload results/RESULTS_LLAMA70B_50CASES.md

Delete results/report.md, results/report_latest.md

Delete EVAL_PROTOCOL.md, RUN_ON_YOUR_MACHINE.md, results_qwen3_4b/, results_qwen3_4b_v2/, results_qwen3_4b_v3/, results_qwen3_4b_v4/, results_qwen3_4b_v5/, results_qwen3_4b_v6/, results_qwen3_4b_v7/, results_real_llm/

Upload RESULTS_ARCHIVE.md

Upload README.md

Upload .gitignore

Upload DESIGN.md

Upload src/pipeline_runner.py

Upload src/config.py

Upload tests/test_constants.py

Upload tests/test_contradiction_checks.py

Upload run_pipeline_batched.py

Upload src/gate.py

Upload src/contradiction_checks.py

Upload RUN_ON_YOUR_MACHINE.md

Upload run_pipeline_batched.py

Upload .env.example

Upload src/llm_client.py

Upload src/config.py

Upload src/gate.py

Upload src/verifier.py

Upload .env.example

Upload tests/test_gate.py

Upload tests/test_contradiction_checks.py

Upload src/diagnose_not_in_evidence.py

Upload src/diagnose_contradictions.py

Upload src/llm_client.py

Upload src/gate.py

Upload src/config.py

Upload src/contradiction_checks.py

Delete results/baseline_normal.jsonl, results/baseline_honesty.jsonl, results/verity_pipeline.jsonl

Update baseline_normal placeholder (raw data in sandbox only)

Add latest 100-case evaluation report (v0.3.2)

Upload results/baseline_normal.jsonl

Upload results/baseline_normal.jsonl

v0.3.2: add EVAL_PROTOCOL.md

v0.3.2: update verifier tests for new parser signature

v0.3.2: report tests — per-category, strict metric, warnings

v0.3.2: add Ph.D./M.D./M.B.A. abbreviation tests

v0.3.2: add dedup priority tests

v0.3.2: update test for new metric key

v0.3.2: add Ph.D., M.D., B.A., B.S., M.S., M.B.A., J.D., D.V.M., D.Phil abbreviations

v0.3.2: dataset version in validation output

v0.3.2: pass case_id to verifier for trace logging

v0.3.2: malformed line tracking, trace logging behind flag

v0.3.2: gate dedup priority — preserve highest-risk label

v0.3.2: per-category breakdown, dataset version in report

v0.3.2: metric denominator fix, among_accepts metric