逃离置信陷阱：扩散大语言模型数学推理的进化解码

中文摘要 AI 辅助

扩散大语言模型（dLLMs）作为自回归大语言模型的有前景替代方案，通过分块渐进式去掩蔽实现高效生成，但其通用性能未必能可靠迁移至数学推理领域，数学推理的正确性依赖于连贯的数值-符号推理轨迹。本研究分析LLaDA 2.0的解码轨迹，发现一种反复出现的扩散置信陷阱：在渐进式分块解码过程中，局部 token 的置信度可能与全局推理正确性错位。我们识别出两类典型失败模式：采样敏感型失败，即存在正确路径但不稳定；采样一致型失败，即重复采样收敛至重复的高置信度但错误的延续。基于此，我们提出进化解码（Evolutionary Decoding），这是一种无训练的测试时扩展框架，将扩散解码视为候选推理状态上的进化过程。该框架结合逐步选择（保留有用的数值-符号信号并抑制重复模式）与分块变异（引入结构化替代方案以逃离错误的高置信度区域）。在多个基准上的实验表明，进化解码相较于基于置信度的解码提升了LLaDA 2.0的性能，实现了更可靠的数学推理。

英文摘要

Diffusion large language models (dLLMs) have emerged as a promising alternative to autoregressive LLMs, offering efficient generation through block-wise progressive unmasking. However, their strong general-purpose performance does not necessarily translate into reliable mathematical reasoning, where correctness depends on preserving coherent numerical-symbolic reasoning trajectories. In this work, we analyze the decoding trajectories of LLaDA 2.0 and identify a recurring diffusion confidence trap: local token confidence can become misaligned with global reasoning correctness during progressive block decoding. Our analysis reveals two representative failure regimes: sampling-sensitive failures, where correct paths exist but are unstable, and sampling-consistent failures, where repeated sampling converges to repetitive high-confidence but incorrect continuations. Motivated by this observation, we propose Evolutionary Decoding, a training-free test-time scaling framework that views diffusion decoding as an evolutionary process over candidate reasoning states. The framework combines step-wise selection, which preserves useful numerical-symbolic signals and suppresses repetitive patterns, with block-wise mutation, which introduces structured alternatives to escape incorrect high-confidence basins. Experiments on multiple benchmarks show that Evolutionary Decoding improves LLaDA 2.0 over confidence-based decoding, leading to more reliable mathematical reasoning.

发表机构

• Australian National University（澳大利亚国立大学）
• Nanyang Technological University（南洋理工大学）
• National University of Singapore（新加坡国立大学）
• Amazon（亚马逊）
• University of Technology Sydney（悉尼科技大学）

机构由 AI 辅助整理，请以论文原文为准。