Masked Language Flow Models

📄 arXiv: 2606.27617v1 📥 PDF

作者: Iskander Azangulov, Kianoosh Ashouritaklimi, Leo Zhang, Simon Vary, Patrick Rebeschini

分类: cs.CL, cs.LG

发布日期: 2026-06-26

备注: Preprint


💡 一句话要点

提出Masked Language Flow Models以解决多步骤推理问题

🎯 匹配领域: 支柱四:生成式动作 (Generative Motion) 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 流式语言模型 多步骤推理 条件生成 掩码机制 连续流

📋 核心要点

  1. 现有的Masked Diffusion Models在少步采样时效率降低,无法充分利用并行生成的优势。
  2. 本文提出Masked Language Flow Models,通过引入掩码和连续流,解决了FLMs在多步骤推理中的局限性。
  3. 在GSM8K和MT-Bench数据集上,MLFMs首次展示了流式语言模型在下游推理任务中的有效性。

📝 摘要(中文)

Masked Diffusion Models(MDMs)在语言生成中展现出快速并行的潜力,但其反向转移在token位置上分解的特性在少步采样时效率降低。Flow Language Models(FLMs)通过学习连续流来克服这一限制,但在复杂的多步骤推理任务中表现不佳。为此,本文提出了Masked Language Flow Models(MLFMs),通过引入掩码和连续随机插值来实现条件生成,并允许预训练的MDMs轻松转化为MLFMs。我们还提出了一种新型采样器,交替进行连续去噪和自信token的离散解掩,以更好地支持多步骤推理。实验结果表明,流式语言模型能够扩展到下游推理和指令跟随任务。

🔬 方法详解

问题定义:本文旨在解决现有流式语言模型在复杂多步骤推理任务中的不足,特别是FLMs在生成过程中需要逐个解码每个token的问题。

核心思路:通过引入掩码机制和连续随机插值,MLFMs能够在部分掩码和干净序列之间建立桥梁,从而实现条件生成,克服FLMs的局限性。

技术框架:MLFMs的整体架构包括掩码机制、连续流的学习和新型采样器,主要模块包括掩码生成、流映射和去噪过程。

关键创新:MLFMs的核心创新在于将掩码引入流式模型中,允许在生成过程中进行条件化,同时通过简单的适配将预训练的MDMs转化为MLFMs。

关键设计:在模型设计中,采用了连续随机插值作为桥梁,损失函数设计考虑了生成的连贯性和准确性,网络结构上则优化了流映射的效率。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果显示,MLFMs在GSM8K和MT-Bench数据集上表现出色,首次证明流式语言模型能够有效解决下游推理和指令跟随任务,性能显著优于传统方法,具体提升幅度未知。

🎯 应用场景

该研究的潜在应用领域包括自然语言处理中的复杂推理任务、对话系统和智能助手等。通过提升流式语言模型的推理能力,未来可能在自动问答、文本生成和指令理解等场景中发挥重要作用。

📄 摘要(原文)

Masked Diffusion Models (MDMs) promise fast, parallel language generation, but their reverse transition factorises across token positions -- an approximation that breaks down in the few-step sampling regime where parallel generation ought to provide the greatest efficiency gains. Flow Language Models (FLMs) sidestep this limitation by learning a continuous flow that transports noise toward clean sequences represented in Euclidean space, inducing a flow map that can be distilled for single-step generation. However, this makes complex tasks requiring multi-step reasoning problematic for FLMs, as FLMs are forced to decode every token during generation. To address this, we introduce Masked Language Flow Models (MLFMs), which incorporate masking into FLMs using a continuous stochastic interpolant to bridge partially masked and clean sequences. This design enables conditional generation via continuous flows and allows pretrained MDMs to be converted into MLFMs through a simple, lightweight adaptation. Leveraging this flexibility, we propose a novel sampler that alternates continuous denoising with the discrete unmasking of confident tokens to better support multi-step reasoning. We evaluate our approach on GSM8K and MT-Bench and find, for the first time, that flow-based language models can be scaled to solve downstream reasoning and instruction-following tasks.