多轮反射掩码激发掩码扩散模型中的推理能力
Multi-Turn Reflective Masking Elicits Reasoning in Mask Diffusion Models
AI总结：
提出反射掩码（RM）方法，通过轻量级后训练使掩码扩散模型具备多轮掩码与去噪能力，实现迭代局部修正，无需架构改变，在文本生成、数独和图像编辑等任务中优于基线。
摘要：
尽管自回归（AR）模型通常通过思维链推理和反思来执行推理，但它们对先前输出的改进仍然依赖于完全顺序生成，即使只需要局部编辑。相比之下，掩码扩散模型（MDMs）中的掩码机制自然支持对先前输出进行显式局部编辑，允许选择性细化而无需丢弃先前答案并从头生成另一个。虽然这一特性更接近人类通过迭代局部修正来纠正错误的方式，但现有的MDMs不支持多轮掩码和去噪。我们提出反射掩码（RM），通过轻量级后训练激发MDMs中这种内在的推理能力。RM提供了一种原生的测试时扩展，其中MDM基于不断演化的上下文迭代地重新审视和修正其先前的输出。为了利用来自先前轮次的见解（如AR推理），我们进一步引入了历史参考，这是一种参数无关的机制，在修正过程中利用中间去噪状态。我们的方法不需要架构改变，并且易于应用于现有的MDMs。在包括文本生成、数独和图像编辑在内的多种任务和模态中，反射掩码始终优于基于标准掩码的基线，并展现出强大的通用性，将RM定位为MDMs上推理的基本原语。
英文摘要：
While reasoning on autoregressive (AR) models is often performed by chain-of-thought reasoning and reflection, their refinement of previous outputs still relies on fully sequential generation, even when only local edits are needed. In contrast, the masking mechanism in Mask Diffusion Models (MDMs) naturally supports explicit local edits on previous outputs, allowing selective refinement without discarding previous answers and generating another from scratch. While this property more closely aligns with how humans correct mistakes by iterative local refinement, existing MDMs do not support multi-turn masking and denoising. We propose Reflective Masking (RM), which elicits such an intrinsic reasoning capability in MDMs via lightweight post-training. RM provides a native test-time scaling, where an MDM iteratively revisits and revises its prior outputs based on evolving context. To exploit insights from previous turns like AR reasoning, we further introduce History Reference, a parameter-free mechanism that leverages intermediate denoising states during revision. Our approach requires no architectural changes and is easily applicable to existing MDMs. Across diverse tasks and modalities, including text generation, Sudoku, and image editing, Reflective Masking consistently outperforms standard masking-based baselines and demonstrates strong generality, positioning RM as a fundamental primitive for reasoning on MDMs.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
