Plan-and-Patch: Diffusion Language Models for Agentic Planning

查看原文 / View Original 编辑 / Edit

Computer Science > Artificial Intelligence arXiv:2610.10786 (cs) [Submitted on 7 Oct 2026] Title: Plan-and-Patch: Diffusion Language Models for Agentic Planning Authors: Syamantak Kumar, Jiang Guo, Hassan Hamad, Hideo Kobayashi, Yi Xiang, Yezhou Yang, Yanjun Qi, Daniele Bonadiman, Jiarong Jiang.

计算机科学 > 人工智能 arXiv:2610.10786 (cs) [提交于 2026 年 10 月 7 日] 标题：Plan-and-Patch：用于智能体规划的扩散语言模型作者：Syamantak Kumar, Jiang Guo, Hassan Hamad, Hideo Kobayashi, Yi Xiang, Yezhou Yang, Yanjun Qi, Daniele Bonadiman, Jiarong Jiang。

Abstract: Planning is increasingly important for long-horizon agents, where successful execution requires coordinating subgoals, tool use, and intermediate outcomes over many steps. Yet assumptions made during planning may be invalidated by the environment, tools may return unexpected results, or actions may fail. Effective agents must therefore not only generate plans, but also revise them.

摘要：规划对于长程智能体而言日益重要，其成功执行需要协调多个步骤中的子目标、工具使用和中间结果。然而，规划期间所做的假设可能会被环境所否定，工具可能会返回意外结果，或者行动可能会失败。因此，有效的智能体不仅必须能够生成计划，还必须能够对其进行修订。

Such revisions often affect only part of a plan, leaving the preceding and subsequent structure intact. Rather than regenerate the entire plan and risk unnecessary changes, repair can regenerate the affected region conditioned on the preserved prefix and suffix. We introduce Plan-and-Patch, a plan-and-act framework in which a diffusion language model (dLLM) generates a structured, program-like plan through parallel unmasking and repairs it by filling in selected regions while keeping the surrounding steps fixed.

此类修订通常仅影响计划的一部分，而保持前后结构完整。与其重新生成整个计划并冒着不必要更改的风险，不如在保留前缀和后缀的条件下，仅重新生成受影响的区域。我们引入了 Plan-and-Patch，这是一个规划与执行框架，其中扩散语言模型 (dLLM) 通过并行去掩码生成结构化的程序式计划，并通过填充选定区域同时保持周围步骤不变来进行修复。

We compare DreamReasoner-8B and Qwen3-8B as diffusion and autoregressive (AR) planners. On Natural Plan without task-specific training, diffusion (53.7%) achieves nearly twice the plan repair success rate of AR (27.0%). After task-specific training on agentic benchmarks, ALFWorld and TextCraft, the planners achieve similar observed success in plan generation, while diffusion reduces mean plan-generation latency by 39-46% relative to AR.

我们比较了 DreamReasoner-8B 和 Qwen3-8B 作为扩散和自回归 (AR) 规划器的表现。在没有特定任务训练的 Natural Plan 上，扩散模型 (53.7%) 的计划修复成功率几乎是 AR (27.0%) 的两倍。在针对智能体基准测试 ALFWorld 和 TextCraft 进行特定任务训练后，这些规划器在计划生成方面取得了相似的观察成功率，而扩散模型相对于 AR 将平均计划生成延迟降低了 39-46%。

Our results show that Plan-and-Patch provides a framework for faster plan generation and effective plan repair in long-horizon agents. Subjects: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG).

我们的结果表明，Plan-and-Patch 为长程智能体提供了更快的计划生成和有效的计划修复框架。学科：人工智能 (cs.AI)；计算与语言 (cs.CL)；机器学习 (cs.LG)。