polyphony · claude-opus-5-5

✓ pass

Pass rate

100%

Cases

Mode

eval

Suite

frankenstein

Env

Started

2026-10-24 06:05 UTC

Duration

—

Cost

—

Tokens in/out

— / —

Commit

ececd15

Branch

main

Source

CI run ↗

Metrics

attribution.accuracy	0.400
attribution.chance	0.333
attribution.score	0.400
continuity.detection_recall	0.500
continuity.false_positive_rate	0
continuity.score	0.500
extraction.f1	0.737
extraction.precision	0.583
extraction.recall	1
extraction.score	0.737
ingestion.score	0.667
retrieval.mrr	0.833
retrieval.precision_at_k	0.815
retrieval.score	0.815

Labels

cadence	on-demand
experiment	w15-post-1
harness_sha	ececd1595c9bfb5b766355795d60833b965ec6a6
judge_provider	groq
judge_self_graded	False
repeat	1
trigger	workflow_dispatch

Compared with the previous run

Baseline 2026-10-23 08:18 UTC · 86% → this run 100% · code diff ↗

✓ newly passing · 1

eval:extraction

± score moved · 2

eval:attribution: 0.60 → 0.40
eval:ingestion: 0.54 → 0.67

cases added / removed

− eval:outline
− eval:prose
− eval:revise

Cases (5)

gated 4/4 passed

eval:continuityscore 0.50✓ pass

score=0.500precision=0.500n_findings=2n_injected=2quote_verbatim=1detection_recall=0.500n_control_findings=0false_positive_rate=0

Output

detection_recall=0.5 fpr=0.0

eval:extractionscore 0.74✓ pass

f1=0.737score=0.737recall=1windows=2precision=0.583f1_selected=0.769recall_selected=0.714precision_selected=0.833principal_fields_rate=1principal_mentions_min=4

Output

F1=0.7368 P=0.5833 R=1.0 predicted=['Soraya', 'Fenwick', 'the being', 'Alina', 'Emeric Vandel', 'Hollis Crane', 'Isolde Sarrow', 'Cosima', 'Rowan Merrow', 'Uncle Thomas', 'the lieutenant', 'the master']

eval:ingestionscore 0.67✓ pass

score=0.667grounded=8avg_chunks=10.33n_characters=12

Output

grounded=8/12 avg_chunks=10.33

eval:retrievalscore 0.81✓ pass

mrr=0.833score=0.815n_queries=9precision_at_k=0.815

Output

precision@k=0.8148 mrr=0.8333

reported 1 reported

eval:attribution0.40 · reported

n=30empty=0score=0.400chance=0.333correct=12accuracy=0.400mean_gap=0.052generated=30

Output

accuracy=0.4 chance=0.3333 n=30