arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.23968cs.LG
收藏

通过变分推断学习掩码离散扩散模型的生成顺序

Learning Generation Orders for Masked Discrete Diffusion Models via Variational Inference

University of Bristol(布里斯托尔大学);Mistral AI

David Fox, Sam Bowyer, Song Liu, Laurence Aitchison, Raul Santos-Rodriguez, Mengyue Yang

AI总结:

本文提出了一种基于变分推断的学习框架,用于优化掩码离散扩散模型的并行生成顺序,通过实验展示了其在生成效率与质量上的竞争力。

摘要:

掩码离散扩散模型(MDMs)是一种有前途的新生成建模方法,能够实现并行令牌生成,因此比自回归方法更高效。然而,实现并行生成与样本质量之间的最佳平衡仍是一个开放问题。当前方法主要通过固定、启发式的并行采样方法来解决这个问题。虽然有一些近期基于学习的方法来解决这个问题,但其从变分推断的角度进行建模仍处于探索阶段。在本文中,我们提出了一种变分推断框架,用于学习MDMs的并行生成顺序。作为我们方法的一部分,我们提出了一个近似后验生成顺序的参数化,这在训练过程中促进了并行性和高效的采样。使用这种方法,我们在GSM8K数据集上进行了初步实验,在高度并行生成的范围内,我们的方法在与启发式采样策略竞争时表现相当。例如,我们的方法在仅平均4个生成步骤的情况下达到了33.1%的准确率,而标准竞争对手方法在同一步骤数中实现了23.7-29.0%的准确率。我们认为进一步的实验和分析将为MDMs的并行生成问题提供有价值的见解。

英文摘要:

Masked discrete diffusion models (MDMs) are a promising new approach to generative modelling, offering the ability for parallel token generation and therefore greater efficiency than autoregressive counterparts. However, achieving an optimal balance between parallel generation and sample quality remains an open problem. Current approaches primarily address this issue through fixed, heuristic parallel sampling methods. There exist some recent learning based approaches to this problem, but its formulation from the perspective of variational inference remains underexplored. In this work, we propose a variational inference framework for learning parallel generation orders for MDMs. As part of our method, we propose a parameterisation for the approximate posterior of generation orders which facilitates parallelism and efficient sampling during training. Using this method, we conduct preliminary experiments on the GSM8K dataset, where our method performs competitively against heuristic sampling strategies in the regime of highly parallel generation. For example, our method achieves 33.1\% accuracy with an average of only only 4 generation steps, compared to 23.7-29.0\% accuracy achieved by standard competitor methods in the same number of steps. We believe further experiments and analysis of the method will yield valuable insights into the problem of parallel generation with 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。

↑