GPT-5.6 Reportedly Solves Long-Standing Statistics Problem
新智元 · wechat · 2026-07-16
This article discusses a long-standing open question in statistics: whether the BH procedure always controls the FDR under the correlated two-sided Gaussian test.
According to the article, Wharton statistics professor Edgar Dobriban and related discussions indicate that GPT-5.6 constructed a counterexample in about 90 minutes, demonstrating that the BH procedure might lose FDR control in this scenario. The article also notes:
- The underlying BH procedure is a classic method for multiple hypothesis testing, widely used in high-throughput science.
- For the past two decades, the academic community has lacked a rigorous proof or counterexample.
- GPT-5.6's approach is described as combining standard asymptotic analysis with numerical certificates, accompanied by a code repository for verification.
- The author interprets this as AI entering the realm of "constructive discovery," potentially shifting scientific research paradigms.
The conclusion reminds readers that the practical impact on high-throughput science still requires further validation.
Related event: GPT-5.6 Reportedly Found a BH Counterexample(8 posts)→
More from Models
- DeepSeek cut KV cache per token from 389KB to 890B in two years as open-closed gap debate rages — teortaxesTex · 2026-10-27
- GPT-6 Astra controls dexterous hands zero-shot at just 5.2% success in 500 trials — stingning · 2026-10-27
- Unverified: fable 5.5 speculated at 4-5T parameters to outscale 1-3T Opus — haider1 · 2026-10-27
- Kimi K3 Is 'Open' Until You Succeed: License Caps Free Use at $20M Revenue — alex_verem · 2026-10-27
- Screenshot shows 'GPT-6 Astra Ultra' in ChatGPT model picker — unverified leak or joke — justalexoki · 2026-10-27
- DeepSeek Safety Eval Finds "Over-Attribution": Model Mistakes Ordinary Tasks for Tests — Much_Boysenberry9124 · 2026-10-27