Evidence report: S_qwen3

Reference: Qwen2.5-1.5B (exact lineage). Probes: 199. Candidate pool: r1_orig, qwen3_32b, gemini, r1, gpt_oss_120b, llama33_70b.

Tier 1: signal gate

Best reasoning-family mean alignment: +0.172 (qwen3) vs gate +0.05. Passed. Matched control: -0.005.

Tier 2: family attribution

Family qwen3 wins 89% of probes (chance 17%); anytime-valid evidence log10 E = 108.25, threshold crossed at probe 14; style probability mass on this family 0.804. Passed.

Tier 3: member / pipeline attribution

Members in pool: qwen3_32b.

Likelihood wins: None; top None (margin p nan).

Style member probabilities: None.

Note: single member in pool; no within-family attribution possible

Disclosure

Candidate pool: r1_orig, qwen3_32b, gemini, r1, gpt_oss_120b, llama33_70b; style classifier pool: r1_orig, r1_openr1, qwen3_32b, gemini, r1, gpt_oss_120b, llama33_70b. A strong preference for a wrong near-relative is what a missing true teacher looks like; the pool must include the suspect's plausible teachers and their pipelines.