arXiv cs.CL· Jaesang Yu, Seoyeon Byun, Yun Wang, Yan Lu, Juho Kim·· 13 小时前AI 评分34
Draft Refinement Language Models：用并行草稿精修替代扩散反向过程做推理
Breaking Diffusion: Parallel Reasoning with Draft Refinement
AI 导读
研究者提出 Draft Refinement Language Models（DRLM），用单个去噪器并行提出修订并评估候选草稿，通过校正自 token 证据得到一次前向即可计算的 cavity energy，仅保留能量更低的候选。
正文 · 原文
Abstract:Drafting an answer allows a reasoning model to reconsider each decision in the context of the others. Yet repeated diffusion denoising does not ensure that individually plausible token updates form a better answer together. We introduce Draft Refinement Language Models (DRLMs), which use a single denoiser to propose parallel revisions and evaluate a candidate draft. We train the denoiser to reconstruct every token from uniformly corrupted drafts, teaching both preservation and repair. Its predictions specify revisions directly, without a diffusion reverse transition. Evaluation requires a different use of these predictions: confidence in an existing token reflects both contextual support and evidence from observing the token itself. Using the known corruption process, we correct for this self-token evidence to estimate how well each token is supported by the prompt and the remaining draft. Aggregating these estimates yields a cavity energy computed in one forward pass, without an additional scoring model. Each round proposes a revision, reads the completed candidate in its revised context, and retains it only if its energy is lower. Across Countdown, Sudoku-Extreme, and Maze-Hard, our method outperforms autoregressive and discrete diffusion baselines. With 6M parameters, DRLM reaches 44.4% on Countdown-5 and 93.9% on Sudoku-Extreme. The results establish complete-draft parallel refinement as an effective alternative to reverse diffusion for complex reasoning.
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.26542 [cs.AI] |
| (or arXiv:2610.26542v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.26542 arXiv-issued DOI via DataCite |
Submission history
From: Jaesang Yu [view email]
[v1]
Mon, 19 Oct 2026 15:02:45 UTC (586 KB)
来源：arXiv cs.CL · arxiv.org