Anonymous Sleuth Shows Nature fMRI Foundation Model Paper Data Was Random, Seed 42
bttyeo · x · 2026-10-26
Blogger Neuroskeptic reports a remarkable case of science sleuthing: an anonymous PubPeer commenter alleges data fabrication in "Towards a general-purpose foundation model for functional MRI analysis" (NeuroSTORM, Cheng Wang et al., published April 2026 in Nature Biomedical Engineering).
The sleuth's plot shows each published datapoint overlaps almost exactly — same values, same order — with data generated by a specific pseudorandom generator, implying the paper's results were randomly generated and that the exact generation pipeline was reverse-engineered.
The recipe: NumPy with np.random.seed(42), generating 400 comparison values per model/metric group — 300 from Normal(center, 3), 80 from Normal(center, 6), 10 from center + Uniform(10, 14), and 10 from center − Uniform(10, 14). Neuroskeptic, who has covered scientific integrity for 20 years, calls it one of the most impressive fraud-detection cases he has seen.
More from Safety
- TenetBench: Clause-Anchored Benchmark Finds LLMs Frequently Violate Ethics Codes — Yegor Anichkov · 2026-10-26
- AI cyber classifier reportedly refuses to help audit Apache code — jd_pressman · 2026-10-26
- 56% of Organizations Cut Cybersecurity Roles Due to AI, Barracuda Report Finds — mikeflache · 2026-10-26
- Deleting browser history doesn't delete Google's account-tied copy — how to wipe it all — nikola_mr64990 · 2026-10-26
- AI researchers lament models refusing vulnerability analysis of old PHP code despite 'use AI to defend' messaging — voooooogel · 2026-10-26
- Even at max Daybreak tier, AI vuln-analysis requests still blocked by cybersecurity guardrails — jd_pressman · 2026-10-26