PIG:用于加速扩散语言模型的并行不一致性引导
PIG: Parallel Inconsistency Guidance for Accelerating Diffusion Language Models
AI总结:
本文提出PIG范式,通过轻量级辅助模型估计并行与顺序生成的不一致性,在多个基准数据集上提升了扩散语言模型的准确率-效率表现。
摘要:
扩散语言模型通过每次前向传播解码多个掩码位置来加速生成。然而,当并行解码轨迹偏离其顺序对应项时,准确率会下降,而顺序对应项是缓慢且昂贵的“基准”,符合语言的联合分布。本文提出了并行不一致性引导(Parallel Inconsistency Guidance,PIG)这一新范式,该范式利用训练好的辅助模型,从单次模型前向传播中直接估计并行与顺序生成之间的不一致性。推理过程中,掩码位置按置信度递减的顺序遍历,每次向并行解码集添加一个标记,直到并行与顺序生成之间的总估计不一致性超过阈值。令人惊讶的是,只需在从OpenCoder、GSM8K-train、MATH-train和MBPP-train中抽取的小型混合数据集上训练一个轻量级的约200万参数辅助模型即可。对于LLaDA-8B Instruct和NemotronDiffusion-3B Instruct,与行业标准Fast-dLLM相比,PIG在GSM8K、MATH500、MBPP、HumanEval和MMLU上拓展了准确率-总前向传播次数(TPF)前沿;与DAPD、KLASS、EB-sampler及其他学习型去掩码策略相比,PIG在LLaDA-8B Instruct上表现出色。
英文摘要:
Diffusion language models accelerate generation by decoding multiple masked positions per forward pass. However, accuracy is known to drop when parallel decoding trajectories deviate from their sequential counterparts, which are slow and expensive "oracles" that adhere to the joint distribution of language. In this paper, we propose Parallel Inconsistency Guidance (PIG), a new paradigm that uses a trained auxiliary model to directly estimate the inconsistency between parallel and sequential generation from a single model forward pass. During inference, masked positions are traversed by decreasing confidence, and tokens are added to the parallel decode set one-at-a-time until the total estimated inconsistency between the parallel and sequential generation exceeds a threshold. Surprisingly, it suffices to train a lightweight ~2M parameter auxiliary model on a small mixture drawn from OpenCoder, GSM8K-train, MATH-train, and MBPP-train. For LLaDA-8B Instruct and NemotronDiffusion-3B Instruct, PIG extends the accuracy-TPF frontier on GSM8K, MATH500, MBPP, HumanEval, and MMLU, when compared to the industry standard Fast-dLLM and measures favorably on LLaDA-8B Instruct against DAPD, KLASS, EB-sampler, and other learned unmasking policies.
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。