---
title: "神经符号具身智能体：将长程家务任务分解为视觉探索与约束符号规划"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.16794v1"
permalink: "https://feed.astralor.com/items/itm_0657259aa53a263d102242a1"
published: "2026-08-17T16:50:59.000Z"
score: 78
channel: "ai"
tags: ["PDDL","VirtualHome","Vision-Language Model","具身智能","神经符号方法","蒙特卡洛树搜索"]
---

# 神经符号具身智能体：将长程家务任务分解为视觉探索与约束符号规划

## AI 点评

> 这篇论文值得关注的地方在于它没有堆更大的模型，而是用神经符号结合让4B小模型也能在具身规划中表现惊艳。对做机器人决策或Agent落地的人来说，这种约束解码加搜索的思路可能是比纯靠推理时计算更省、更可控的路径。

## 速览

语言和视觉语言模型能生成看似合理的具身规划，但无法保证可执行性，输出可能违反环境动态或作用于错误实体。本文提出一种神经符号智能体，将长程家务任务分解为任务导向的视觉探索和约束符号规划。第一阶段，视觉语言模型与探索机制从第一视角观察和交互中获取目标相关谓词和实例绑定，生成符号初始状态；第二阶段，PDDL转移模型将解码限制为可应用的行动，蒙特卡洛树搜索用与领域无关的规划启发式评估可执行延续。由此生成的规划在转移模型下天然可执行，在VirtualHome和ALFWorld上，4B-27B开源模型成功率均超90%，最小智能体大幅超越27B直接视觉策略。约束与搜索互补，在ALFWorld中单独使用任一方法只能解决不到三分之一任务，组合后解决超95%。该方法生成的token和模型可见图像远少于直接交互，残余失败集中在状态获取而非规划生成。

## 关键点

- 神经符号智能体将长程家务任务分解为视觉探索和符号规划两阶段，第一阶段生成符号初始状态，第二阶段用PDDL约束解码和MCTS搜索保证可执行性。
- 在VirtualHome和ALFWorld上，4B-27B开源模型均超90%成功率，最小智能体大幅优于27B直接视觉策略。
- 约束与搜索互补：在ALFWorld中单独使用任一方法只能解决不到三分之一任务，组合后解决超95%。
- 方法比扩展思维生成少数倍的token，比直接交互需要少得多的视觉图像，残余失败主要来自状态获取而非规划生成。

## 全文

Language and vision-language models generate plausible embodied plans but do not guarantee executability, as their outputs can violate environment dynamics or act on incorrectly grounded entities. We present a neurosymbolic agent that factors long-horizon household tasks into task-directed visual exploration and constrained symbolic planning. In the first phase, a vision-language model and exploration harness acquire goal-relevant predicates and instance bindings from egocentric observations and grounded interactions, producing a symbolic initial state. In the second, a PDDL transition model restricts decoding to tokens that extend applicable actions. Monte Carlo tree search then evaluates executable continuations using a domain-independent planning heuristic. The resulting plans are executable by construction under the transition model, with transfer to the environment conditioned on correct visual grounding. On VirtualHome and ALFWorld, open 4B-27B models exceed 90% success in both environments, and our smallest agent substantially outperforms a 27B direct visual policy in each. Constraints and search prove complementary rather than interchangeable: in ALFWorld either alone solves under a third of tasks, whereas their combination solves over 95%. The method also uses several times fewer generated tokens than extended thinking and far fewer model-visible images than direct interaction, and residual failures localize to state acquisition rather than plan generation without any specialized training.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_0657259aa53a263d102242a1