arXiv:2610.25596 · v2

Cognitive Maps in Large Language Models

首次提交 2026-10-19 08:02

版本更新 2026-11-06 22:33

雷达观测 2026-11-09 13:22

摘要

It has long been proposed that animals construct cognitive maps to support flexible behavior in novel environments. We tested whether large language models form analogous internal representations using relational reasoning tasks and causal "mosaic" experiments. Models beyond the ten-billion-parameter scale inferred unstated relationships and generalized to synthetic entity names. Across models and tasks, models formed abstract representations of relational structure, additively combining them with representations of entity identity. Relationships were primarily stored as pairwise links and composed retrospectively, rather than prospectively encoded in a global coordinate system. A sparse, distributed set of key-value pairs mediated binding; perturbing them implanted "false memories" that produced predictable behavioral changes. Our findings reveal a conserved organization analogous to a cognitive map supporting flexible and systematic reasoning in language models.

作者备注：Changes in v2: Revised Discussion, minor edits throughout, and added a summary figure (Fig. 6). Results unchanged