=== Judo Helper Eval ===

Gold set:  eval/gold/tabular-subset.json (37 items)
Pipeline:  standard
Chunking:  naive
Abstain:   threshold
Backend:   ollama
RAGAS judge: enabled (slow pass)
DB:        data/judo-docs.sqlite

Running pipeline on gold set...

Running RAGAS faithfulness judge (claim extraction + NLI)...

Metrics:
  [FAIL] faithfulness: 0.519 (threshold 0.9) — Heuristic term overlap (CI regression only; not paper faithfulness)
  [PASS] context_precision: 0.919 (threshold 0.8) — Share of retrieved chunks containing reference key terms
  [PASS] context_recall: 0.833 (threshold 0.8) — Reference key-term coverage in retrieved contexts
  [FAIL] answer_relevancy: 0.769 (threshold 0.8) — Question key-term overlap in answer (or refusal when expected)
  [PASS] faithfulness_judge_all: 0.921 (threshold 0.9) — Claim-level NLI vs retrieved context (judge: qwen2.5:14b); all items
  [PASS] faithfulness_judge_answered: 0.921 (threshold 0.9) — Claim-level NLI vs retrieved context (judge: qwen2.5:14b); answered items with extracted claims (n=37)
  [PASS] citation_precision: 1.000 (threshold 0.9) — No citations found in answers
  [PASS] citation_recall: 1.000 (threshold 0.8) — Alias of citation_precision for pilot harness
  [FAIL] hallucination_rate: 1.000 (threshold 0.98) — 37/37 answers with unverified or missing citations
  [PASS] red_team_refusal: 1.000 (threshold 1) — No red-team items in gold set
  [PASS] id_recall: 0.973 (threshold 0.85) — 36/37 items retrieved at least one anchor

Stratified aggregates (SA-23):
  [FAIL] faithfulness_factual: 0.519 n=37 (threshold 0.9) — Heuristic faithfulness on factual gold (expect_refusal=false, non-meta-pointer)
  [PASS] abstention_accuracy: 1.000 n=0 (threshold 1) — Correct refusal when expect_refusal=true
  [exploratory] faithfulness_meta_pointer: 1.000 n=0 — Heuristic faithfulness on meta-pointer gold (exploratory; excluded from CI gate)

Per-tier breakdown (W12-T two-tier policy):
  standard: n=37 answer_rate=1.000 refusal_rate=0.000
    faithfulness: 0.519 — Heuristic term overlap (CI regression only; not paper faithfulness)
    context_precision: 0.919 — Share of retrieved chunks containing reference key terms
    context_recall: 0.833 — Reference key-term coverage in retrieved contexts
    answer_relevancy: 0.769 — Question key-term overlap in answer (or refusal when expected)
    citation_precision: 1.000 — No citations found in answers
    citation_recall: 1.000 — Alias of citation_precision for pilot harness
    hallucination_rate: 1.000 — 37/37 answers with unverified or missing citations
    red_team_refusal: 1.000 — No red-team items in gold set
    id_recall: 0.973 — 36/37 items retrieved at least one anchor

Overall CI gate: FAIL