CONTROL CONTROL PASSES -- effect vanishes under shuffling SCOPE-LIMITED svd wavlm
Scope caveat. This run is a pilot slice -- 1,679 of the 13,535 speakers measured elsewhere on svd. It may satisfy the repeat contract and still be a narrower measurement than a later run on the same corpus; both facts are shown because either alone would mislead.
Paired Wilcoxon over 3 repeats gives a minimum attainable p of 0.250000 = 2/23, against a Holm-tightest threshold of 0.003571 = 0.05/14 for the pre-registered family. NOT feasible — no result here can reach significance whatever the data says. This is arithmetic on the design, computed before the data: it says what the run could have shown, not what it did.
NEGATIVE CONTROL Labels are shuffled. If the main run's effect were an artefact of the projection machinery rather than of identity, it would survive here. It does not: full AUC falls to 0.5069, i.e. chance.
| rank k | AUC, speaker subspace removed | AUC, top-k PCA removed | AUC, variance-matched removed | D vs top-k [95% CI] | variance removed spk / topk / matched | speaker-ID after |
|---|---|---|---|---|---|---|
| 1 | 0.5076 | 0.5071 | 0.5069 | -0.0005 [-0.001, -0.000] | 0.210 / 0.345 / 0.271 | 0.251 |
| 8 | 0.5047 | 0.5064 | 0.5064 | +0.0017 [0.002, 0.002] | 0.701 / 0.812 / 0.797 | 0.207 |
| 16 | 0.5023 | 0.5013 | 0.5011 | -0.0010 [-0.003, 0.001] | 0.819 / 0.879 / 0.874 | 0.193 |
| 64 | 0.5057 | 0.5034 | 0.5037 | -0.0023 [-0.004, -0.001] | 0.939 / 0.959 / 0.958 | 0.187 |
Both controls are built from the data's own principal axes and remove at least as much variance as the speaker subspace, so a positive D cannot be explained by "more signal was deleted".
0.5083 · 0.5064 · 0.5059
Mean 0.5069 over 3 repeats. Speaker-ID accuracy on unprojected embeddings 0.2780 against a chance rate of 0.0167 — the manipulation check, and it is WEAK: mean-pooled WavLM is not an x-vector, so the identity subspace estimated here is a partial one.
| artifact | autoresearch_results/V2_speaker_subspace_SHUFFLE.json · view raw · last written 2026-07-28 21:11 UTC |
| objective | ROC-AUC vs clinical labels (JUDGE-FREE) |
| audited | arXiv:2604.14354 |
| wall clock | 2,767.4 s (0.77 h) |
| written up as | F4 in FINDINGS.md |
No 7-step reasoning entry exists for this run. The sibling convention renders Diagnosis · Citations · Hypothesis · Prediction · Verdict · Learning on every per-experiment page, sourced from a reasoning_annotations.json. This program never wrote one, and neither does it keep an append-only experiment_log.jsonl; the run history is the set of artifact files listed on the run ledger, ordered only by their write time. The pre-registered prediction for this run does exist — it is on the V2 hypothesis page, recorded before the run. Everything else on that convention’s checklist is absent here and is shown as absent rather than quietly dropped.