Run 2026-09-14_x6_strict_prompt_v02

Preregistered decision rule (PREREGISTRATION.md section 5)

• H1 (planted recall 0/0 vs decoy false positives 0/0, Fisher one-sided): p = 1.0000 -> FAIL
• H2 (sampler enrichment, hypergeometric): -> FAIL
• H3: not computable (missing S0 or B4)

Decision: NULL (signal requires H1 and H2 both passing).

Run summary

Total measured cost: $9.930771

Models: - cards: claude-haiku-4-5-20251001 (alias haiku) - critic: claude-haiku-4-5-20251001 (alias haiku) - critic_haiku: claude-haiku-4-5-20251001 (alias haiku) - generator: claude-sonnet-5 (alias sonnet) - match_gold: claude-haiku-4-5-20251001 (alias haiku)

Exploratory finds (non-planted survivors, manual review only): 0

Oracle recall on planted: 0.0% [0, 0]; permutation p = 1.0000

T1. Corpus

notes	cards	cards_per_note	bridges	decoys	leakage_failures	corpus_sha256
96	491	5.11	16	12	13	f8a6e9e33577

T2. Sampler enrichment; base rate 0.38% of 119205 cross-note card pairs are planted

arm	draws	planted_hits	distinct_bridges	expected_hits	p_hypergeom

T3. Oracle set: generator and critic without the sampler; mean cosine to gold on planted = 0.000

S0 group	units	NONE rate	non-NONE rate	survivor rate (post critic + dupgate)
recall on planted (gold match)	0			0.0% [0, 0]

T4. Per arm: NONE rate, critic kill rate, survivors, recovered planted bridges, measured cost

arm	units	NONE	critic kill	dup	survivors	bridges reachable	bridges recovered	recall	cost usd	usd per recovered
B4	55	47.3% [35, 60]	27.6% [15, 46]	0	21	16	6	37.5% [18, 61]	6.4149	1.0692

T5. Label-permutation null over the S0 oracle set

statistic	observed	expected under null	p	shuffles
survivors among planted S0 units	0	0.0	1.0000	2000

T8. Recall after critic, per critic (same generations, before the duplicate gate)

critic	model id	correct recoveries	surviving	killed	decoy answers	decoy surviving	kill rate on random
haiku	claude-haiku-4-5-20251001	0	0	0	0	0	0.0% [0, 0]

T9. Filler abstention under the strict single-note prompt: the single-note arm split by note kind. The filler row is the v0.3 validity precondition; bridge and decoy rows for comparison

note kind	units	answered	NONE	errors	answer rate
bridge	32	14	18	0	43.8% [28, 61]
filler	23	15	8	0	65.2% [45, 81]