arXiv：cs.CL· Minhae Oh, Nakyung Lee, Jungwoo Lee·· 4 小时前
DCGC：用掩码扩散模型做草稿条件全局纠错，提升复杂推理
DCGC: Draft-Conditioned Global Correction for Complex Reasoning with Masked Diffusion Models
AI 导读
DCGC 是一个基于 Masked Diffusion Model（MDM）的全局纠错框架，用上游求解器的不完美解答草稿作为辅助上下文，结合任务特定 SFT 与推理时机制 Dynamic Dual-CFG。该机制分离仅问题与问题-草稿联合分支，并按相对置信度差缩放草稿条件残差。在数学、代码与知识推理基准上，DCGC 优于标准采样与更简单的 CFG 变体，并能纠正低共识上游输出以提升全测试集准确率。
正文
Abstract:Correcting flawed reasoning traces remains a significant challenge for Large Language Models (LLMs), whose autoregressive generation can propagate early mistakes into subsequent reasoning. We introduce DCGC, a Masked Diffusion Model (MDM) framework for global correction that uses an imperfect solution draft from an upstream solver as auxiliary context. DCGC combines task-specific Supervised Fine-Tuning (SFT) with a novel inference-time mechanism called Dynamic Dual-CFG. This mechanism separates problem-only and joint problem-draft branches and scales the draft-conditioned residual using a relative confidence gap. Across math, code, and knowledge reasoning benchmarks, DCGC outperforms standard sampling and simpler CFG variants, with additional results suggesting transfer to different diffusion backbones. In test-time setting where ground-truth failure labels are unavailable, DCGC improves full test set accuracy by correcting low-consensus upstream outputs, highlighting its utility as a verifier-free global correction module for difficult reasoning instances.
| Comments: | 21 pages, 3 figures, 12 tables |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.25428 [cs.CL] |
| (or arXiv:2608.25428v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2608.25428 arXiv-issued DOI via DataCite |
Submission history
From: Nakyung Lee [view email]
[v1]
Wed, 26 Aug 2026 06:35:26 UTC (344 KB)
[v2]
Wed, 14 Oct 2026 07:35:27 UTC (344 KB)
来源：arXiv：cs.CL · arxiv.org