Search papers, labs, and topics across Lattice.
This paper introduces DCGC, a Masked Diffusion Model framework designed to correct flawed reasoning in Large Language Models by utilizing imperfect solution drafts as auxiliary context. By integrating task-specific Supervised Fine-Tuning with a novel inference mechanism called Dynamic Dual-CFG, DCGC effectively distinguishes between problem-only and joint problem-draft branches, enhancing the model's ability to correct errors. The results demonstrate that DCGC significantly improves accuracy across various reasoning benchmarks, particularly in scenarios lacking ground-truth failure labels, establishing its role as a powerful verifier-free correction tool.
DCGC corrects flawed reasoning in LLMs by leveraging imperfect drafts, leading to significant accuracy improvements in complex reasoning tasks.
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.
