Continuous Interaction Diffusion: A Diffusion-Native Runtime for Asynchronous Tool-Augmented Reasoning
作者: Yuhang Cao
分类: cs.AI
发布日期: 2026-08-11
备注: 15 pages, 5 figures, 1 table
💡 一句话要点
提出连续交互扩散以解决工具增强推理中的限制问题
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 扩散语言模型 工具增强推理 信息检索 模型架构 去噪过程 认知张量 实时计算
📋 核心要点
- 现有的扩散语言模型在工具使用上受到限制,无法有效整合外部信息与推理过程。
- 提出连续交互扩散(CID),通过将工具交互融入迭代去噪,提升模型的推理效率。
- 本文未提供实证性能数据,主要集中在只读工具的架构设计与评估协议的定义上。
📝 摘要(中文)
大型语言模型越来越依赖外部工具以获取最新信息、执行计算和与外界互动。自回归模型的工具使用与生成过程自然契合,但扩散语言模型(dLLMs)通过并行地反复细化输出的多个部分进行推理,这种停止-恢复的交互模式显得过于限制。为此,本文提出了连续交互扩散(CID),一种将工具交互整合到迭代去噪中的扩散原生模型-运行时架构。CID设计了一个只读事实通道、一个由类型认知张量表示的思维通道和一个显示通道,使得信息需求可以在文本或JSON调用完全序列化之前出现,从而在去噪过程中启动外部读取。返回的结果被投影到不断演变的思维状态中,可以修正早期的认知和显示区域。CID旨在更早地暴露证据,重叠工具延迟与模型计算,减少重复的外部工作,并在新证据到达后保留有用的计算。
🔬 方法详解
问题定义:本文旨在解决扩散语言模型在工具使用中的限制,现有方法在推理过程中无法灵活整合外部信息,导致效率低下。
核心思路:提出连续交互扩散(CID),通过在去噪过程中实现工具交互,允许信息需求在调用完成前就被满足,从而提高推理的灵活性与效率。
技术框架:CID架构包括三个主要通道:只读事实通道、思维通道(由类型认知张量表示)和显示通道。信息可以在去噪过程中被动态读取和更新。
关键创新:CID的创新在于将工具交互与去噪过程结合,允许模型在推理过程中并行处理外部信息,避免了传统方法的停顿与恢复。
关键设计:CID设计了持久绑定机制,以重用静态结果并在需要时刷新变化的源,同时定义了训练目标和评估协议,以确保任务质量和端到端效率。
🖼️ 关键图片
📊 实验亮点
尽管本文未提供具体的实验结果,但CID的设计理念旨在通过更早地暴露证据和重叠工具延迟与模型计算,显著提高推理效率和任务准确性,减少不必要的外部调用。
🎯 应用场景
该研究的潜在应用领域包括智能助手、自动化决策系统和实时信息检索等。通过更高效的工具交互,CID可以在需要快速响应和准确推理的场景中发挥重要作用,提升用户体验和系统性能。
📄 摘要(原文)
Large language models increasingly rely on external tools to access up-to-date information, perform computation, and interact with the outside world. For autoregressive models, tool use naturally fits the generation process: the model emits a tool call, waits for the result, and then continues generating. Diffusion language models (dLLMs), however, reason by repeatedly refining many parts of their output in parallel, making this stop-and-resume interaction pattern unnecessarily restrictive. It can force tool decisions before the model's reasoning has stabilized, delay useful observations until a discrete call finishes, and introduce redundant refinement and tool execution, potentially hurting both task accuracy and inference efficiency. We introduce Continuous Interaction Diffusion (CID), a diffusion-native model--runtime architecture that integrates tool interaction into iterative denoising. CID separates a model-read-only fact channel, a thought channel represented by a Typed Cognitive Tensor, and a display channel. Information needs can emerge before a textual or JSON call is fully serialized, allowing perceptual bindings to launch external reads while denoising continues. Returned results are projected into the evolving thought state and can revise earlier cognition and display regions. Persistent bindings reuse static results without repeated external execution and refresh changing sources when needed. CID is designed to expose evidence earlier, overlap tool latency with model computation, reduce duplicate external work, and preserve useful computation after new evidence arrives. We formalize the architecture, runtime, and training objectives, and define an evaluation protocol for task quality and end-to-end efficiency. This first paper focuses on read-only tools and makes no empirical performance claims.