[2603.26323] From Human Cognition to Neural Activations: Probing the Computational Primitives of Spatial Reasoning in LLMs

Computer Science > Computation and Language arXiv:2603.26323 (cs) [Submitted on 27 Mar 2026] Title:From Human Cognition to Neural Activations: Probing the Computational Primitives of Spatial Reasoning in LLMs Authors:Jiyuan An, Liner Yang, Mengyan Wang, Luming Lu, Weihua An, Erhong Yang View a PDF of the paper titled From Human Cognition to Neural Activations: Probing the Computational Primitives of Spatial Reasoning in LLMs, by Jiyuan An and 5 other authors View PDF HTML (experimental) Abstract:As spatial intelligence becomes an increasingly important capability for foundation models, it remains unclear whether large language models' (LLMs) performance on spatial reasoning benchmarks reflects structured internal spatial representations or reliance on linguistic heuristics. We address this question from a mechanistic perspective by examining how spatial information is internally represented and used. Drawing on computational theories of human spatial cognition, we decompose spatial reasoning into three primitives, relational composition, representational transformation, and stateful spatial updating, and design controlled task families for each. We evaluate multilingual LLMs in English, Chinese, and Arabic under single pass inference, and analyze internal representations using linear probing, sparse autoencoder based feature analysis, and causal interventions. We find that task relevant spatial information is encoded in intermediate layers and can causally influence behavi...