# 可解释世界模型的神经符号学习

> 原文链接：https://robotics.hellobike.com/papers/2026-10-14-2610-19955
> 频道：papers | 更新日期：2026-10-14 | 标签：arxiv, cs.RO
> 核心摘要：提出CGWM，用时序逻辑知识学习可解释概念并以可微推理规划，在两个游戏环境验证更稳健。

# 可解释世界模型的神经符号学习

::: v-pre
> 原文链接：[Neurosymbolic Learning of Interpretable World Models](https://arxiv.org/abs/2610.19955)（arXiv 2610.19955 · 发布日期：2026-10-14）

**Neurosymbolic Learning of Interpretable World Models**

提出CGWM，用时序逻辑知识学习可解释概念并以可微推理规划，在两个游戏环境验证更稳健。

## 核心要点

- 现有世界模型多学习黑盒表示，难以纳入关于世界的先验知识。
- CGWM从观测中学习可解释概念表示，并用时序逻辑刻画有效状态与动态关系。
- 将规划表述为可微推理，通过梯度上升优化动作以满足目标。
- 在两个游戏环境中，相比纯神经和神经符号基线，概念扎根更准确，分布内外规划更稳健。

**验证范围与局限**：验证范围为两个游戏环境。

## 英文摘要 (Abstract)

World models aim to learn compact representations of an environment from observations, enabling agents to predict future states and plan their behaviour. However, most existing approaches learn black-box representations, with limited ability to incorporate prior knowledge about the world. We introduce Concept-Grounded World Models (CGWMs), a neurosymbolic world model design that instead learns interpretable concept representations. We learn these concepts from observations using explicit and complete knowledge, expressed in temporal logic, that specifies the valid world states and the temporal relations governing their dynamics. Furthermore, we directly use this knowledge by formulating planning as differentiable reasoning, optimising the agent's actions by gradient ascent to satisfy a goal. We experimentally compare CGWM with a purely neural baseline and a neurosymbolic baseline on two game environments. We find that CGWMs ground concepts much more accurately, even over long horizons, resulting in significantly more robust planning both in and out of distribution.

作者：Riccardo Andreoni, Massimiliano Ronzani, Chiara Ghidini, Alessandro Daniele, Emile van Krieken
:::

## 原始参考来源 (Sources)

- [arXiv 2610.19955：Neurosymbolic Learning of Interpretable World Models](https://arxiv.org/abs/2610.19955)