97% of 210 biomedical AI studies used invalid statistical tests for model comparison
bttyeo · x · 2026-05-21
- New bioRxiv preprint led by Tianchu Zeng and B.T. Thomas Yeo (with Konrad Kording, Danilo Bzdok, Thomas Nichols et al.) meta-analyzes 210 biomedical AI studies that statistically compared models under cross-validation.
- 97% used invalid statistical tests — spurious model comparisons are widespread in biomedical AI.
- Framework: a trained model is the product of data + algorithm; fixing the algorithm and swapping data modality or preprocessing lets performance differences quantify the contribution of each.
More from Research
- AnyGameRenderer turns AI agent-built graybox game footage into high-fidelity visuals — AlayaLab · 2026-10-23
- LayerEditor distills edit priors to turn image edits into reusable RGBA layers — jingdong1 · 2026-10-23
- F2G-Pose: real-time category-level pose estimation from a single RGB-D image — sii-research · 2026-10-23
- KAIST's FRED lifts robot policy success from 19.4% to 50% via self-distilled future video representations — kaist-ai · 2026-10-23
- Memory-bound at batch 1, compute-bound at batch N: how GPUs actually run LLMs — techNmak · 2026-10-23
- NVIDIA's ArtiFixer fixes 3D reconstruction artifacts with 8B model, 3-4x faster than 14B — ZGojcic · 2026-10-23