Persistent-memory denoiser hits 99.9% on Sudoku-Extreme with under 250k parameters
LucaAmb · x · 2026-09-12
Researchers added a persistent memory state and removed timestep conditioning from a denoising model, yielding an anytime iterative solver. Since classic sudoku solvers need memory to track possible states, persistent memory between denoising steps proves key: with fewer than 250k parameters the model scores 99.9% on Sudoku-Extreme and 98.3% on Maze-Unique — a new take on reasoning with atypical diffusion models.
Related event: Denoiser with Persistent Memory Solves Sudoku-Extreme at 99.9%(2 posts)→
More from Research
- Curie: a from-scratch 17B model designed to run from SSD, 33 tokens/s on one CPU core — Just_Vugg_PolyMCP · 2026-09-12
- Fortnow: Persistent Writable Memory Could Enable Recursive Self-Improvement — fortnow · 2026-09-12
- Researcher wires a fruit fly brain to ChatGPT with 16-word outputs passing 100% held-out tests — neuroecology · 2026-09-12
- Reasoning Vision-Language Model Improves Multilingual Entity Linking — neulab · 2026-09-12
- SkySynth co-evolves formal proofs with code: 2.3x faster KV stores, 2.2x vLLM throughput — CShorten30 · 2026-09-12
- Evals are increasingly unreadable: overly strict hidden tests mislabel good answers — trq212 · 2026-09-12