前缀去噪一致性：扩散语言模型的测试时验证方法

中文摘要 AI 辅助

扩散语言模型（DLMs）近年来竞争力不断提升，在部分任务上甚至超越了自回归（AR）模型。与AR模型不同，DLMs通过迭代去噪生成输出，无需遵循从左到右的顺序。为进一步提升DLMs性能，我们提出PDC（前缀去噪一致性，Prefix-Denoising Consistency），一种适用于DLMs的测试时自验证方法。PDC利用DLMs在前缀条件再生下的独特测试时信号：正确轨迹比错误轨迹更稳定、可复现。具体而言，给定初始生成样本，PDC将句子在中间位置拆分，基于固定前缀再生剩余 token。在数学推理和常识推理基准测试中，PDC持续提升初始样本质量，在计算受限的对比中优于独立生成结果，且对不同的去掩策略和参数设置具有鲁棒性。这些结果表明，前缀条件再生是一种适用于DLMs的有效测试时验证基础模块。

英文摘要

Diffusion Language Models (DLMs) have recently become increasingly competitive with autoregressive (AR) models, and even outperform them on certain tasks. Unlike AR models, DLMs produce output through iterative denoising without a left-to-right order. To further improve the performance of DLMs, we introduce PDC (\emph{Prefix-Denoising Consistency}), a test-time self-verification method for DLMs. PDC exploits a distinctive test-time signal in DLMs under prefix conditioned regeneration, correct trajectories are more stable and reproducible than incorrect ones. Concretely, given an initially generated sample, PDC splits the sentence at an intermediate position and regenerates the remaining tokens conditioned on the fixed prefix. Across mathematical reasoning and commonsense reasoning benchmarks, PDC consistently improves upon the initial sample, outperforms independent generations under a computational constrained comparison, and is robust to different unmasking strategies and parameter settings. These results highlight prefix-conditioned regeneration as an effective DLM-specific primitive for test-time verification.

发表机构

• MBZUAI（穆罕默德·本·扎耶德人工智能大学）
• Nagoya University（名古屋大学）
• RIKEN AIP（理化学研究所人工智能项目）

机构由 AI 辅助整理，请以论文原文为准。