Sampling the SK Model at β<1: New Polynomial-Time Algorithm Closes a 2022 Open Problem
canondetortugas · x · 2026-09-29
A new arXiv paper (2609.30590) by Lee, Sandhu, and Shi gives a polynomial-time algorithm to sample the Sherrington-Kirkpatrick model's Gibbs measure with vanishing TV error at any inverse temperature β<1, resolving a question open since the 2022 denoising diffusion proposal. The method combines algorithmic stochastic localization with rejection sampling over path-space via Jarzynski's equality, replacing global regularity requirements with local regularity established via Celentano's proof of local strong convexity of the TAP free energy.
More from Research
- Anthropic: Infrastructure Config Alone Swings Agentic Coding Benchmarks by 6 Points — giansegato · 2026-09-29
- GoFish demo: hierarchical chart construction lets you attach animations directly to chart structure — arvindsatya1 · 2026-09-29
- Eval author warns against over-indexing on benchmarks; Anthropic finds infra noise can swing scores 6 points — giansegato · 2026-09-29
- Claim: Transformers can now be pretrained with zeroth-order optimization, no backprop — teortaxesTex · 2026-09-29
- DR-GEM: Self-Supervised Framework Fixes Dimensionality Reduction Bias Against Rare Cell Types — anshulkundaje · 2026-09-29
- AI agents claim Collatz breakthrough: positive proportion of numbers reach 1, Lean-formalized — AlexKontorovich · 2026-09-29