arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2604.10690cs.AI
收藏

大型语言模型构建空间世界模型了吗?基于网格世界迷宫任务的证据

Do LLMs Build Spatial World Models? Evidence from Grid-World Maze Tasks

University of Notre Dame(圣母大学);Columbia University(哥伦比亚大学);MQube Cognition(MQube认知实验室)

Weijiang Li, Yilin Zhu, Rajarshi Das, Parijat Dube

AI总结:

本文通过迷宫任务评估LLM的空间理解能力,发现其空间推理依赖于表示形式而非任务类型,表明LLM缺乏稳健的空间世界模型。

摘要:

基础模型在多样化任务中表现出色,但其构建内部空间世界模型的能力仍不明确。我们通过迷宫任务系统评估了大语言模型的空间理解能力,发现Gemini-2.5-Flash、GPT-5-mini、Claude-Haiku-4.5和DeepSeek-Chat在不同表示形式下空间推理能力差异显著。使用链式思维提示,Gemini在小型迷宫中准确率高达80-86%,但使用视觉网格格式时准确率骤降至16-34%,差异达2-5倍,表明空间推理依赖于表示形式而非任务类型。进一步通过顺序接近问题和组合距离比较探测空间理解,尽管模型在推理轨迹中达到96-99%的语义覆盖,但未能利用此理解进行一致的空间计算,表明模型将每个问题独立处理而非构建累积空间知识。基于迷宫求解任务的结果表明,LLM未发展出稳健的空间世界模型,而是表现出依赖于表示形式和提示的推理能力,仅在狭窄条件下有效。这些结果对部署基础模型于需要空间抽象的应用具有关键影响。

英文摘要:

Foundation models have shown remarkable performance across diverse tasks, yet their ability to construct internal spatial world models for reasoning and planning remains unclear. We systematically evaluate the spatial understanding of large language models through maze tasks, a controlled testing context requiring multi-step planning and spatial abstraction. Across comprehensive experiments with Gemini-2.5-Flash, GPT-5-mini, Claude-Haiku-4.5, and DeepSeek-Chat, we uncover significant discrepancies in spatial reasoning that challenge assumptions about LLM planning capabilities. Using chain-of-thought prompting, Gemini achieves 80-86% accuracy on smaller mazes (5x5 to 7x7 grids) with tokenized adjacency representations, but performance collapses to 16-34% with visual grid formats, which is a 2-5x difference, suggesting representation-dependent rather than format-invariant spatial reasoning. We further probe spatial understanding through sequential proximity questions and compositional distance comparisons. Despite achieving 96-99% semantic coverage in reasoning traces, models fail to leverage this understanding for consistent spatial computations, indicating that they treat each question independently rather than building cumulative spatial knowledge. Our findings based on the maze-solving tasks suggest that LLMs do not develop robust spatial world models, but rather exhibit representation-specific and prompting-dependent reasoning that succeeds only under narrow conditions. These results have critical implications for deploying foundation models in applications requiring spatial abstraction.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑