Recurrent denoisers: adding a persistent hidden state makes diffusion an anytime solver
burkov · x · 2026-09-11
The paper explores iterative solvers that improve simply by running longer at inference. While standard diffusion carries information only in its noisy state, the authors test whether adding a persistent hidden state to a timestep-free diffusion denoiser yields an anytime solver that refines solutions without explicit time conditioning or fixed rollout lengths.
Recurrent denoisers — looped Transformers for Sudoku-Extreme and local convolutional blocks for Maze-Unique — were trained on millions of unique-solution puzzles using 4-step truncated back-prop segments and an annealed noise schedule. At inference, the same shared update runs for thousands of steps with the hidden state carried forward, so accuracy grows with compute.
More from Research
- Google's ToolGrad generates tool-use datasets answer-first, hitting near 100% pass rate — DuRuofei · 2026-09-11
- Multi-agent evals still undecided, but colocated async RL training is catching on — stochasticchasm · 2026-09-11
- Does DeepSeek V4.1-Flash's SWA Bounded Replay sacrifice recall to save KV cache memory? — Top-Handle-5728 · 2026-09-11
- Preprint quantifies proteomics data leakage: no real signal still yields AUCs near 0.8 — bttyeo · 2026-09-11
- Santa Fe Institute opens 2027 Complexity Postdoctoral Fellowships, deadline Sep 30, 2026 — yoavartzi · 2026-09-11
- ego2wrist: Faking robot wrist-camera views from egocentric video, then testing if they work — chris_j_paxton · 2026-09-11