Linear Spatial World Models Emerge in Large Language Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915321682591744 |
|---|---|
| author | Tehenan, Matthieu Moya, Christian Bolivar Long, Tenghai Lin, Guang |
| author_facet | Tehenan, Matthieu Moya, Christian Bolivar Long, Tenghai Lin, Guang |
| contents | Large language models (LLMs) have demonstrated emergent abilities across diverse tasks, raising the question of whether they acquire internal world models. In this work, we investigate whether LLMs implicitly encode linear spatial world models, which we define as linear representations of physical space and object configurations. We introduce a formal framework for spatial world models and assess whether such structure emerges in contextual embeddings. Using a synthetic dataset of object positions, we train probes to decode object positions and evaluate geometric consistency of the underlying space. We further conduct causal interventions to test whether these spatial representations are functionally used by the model. Our results provide empirical evidence that LLMs encode linear spatial world models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2506_02996 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Linear Spatial World Models Emerge in Large Language Models Tehenan, Matthieu Moya, Christian Bolivar Long, Tenghai Lin, Guang Artificial Intelligence Large language models (LLMs) have demonstrated emergent abilities across diverse tasks, raising the question of whether they acquire internal world models. In this work, we investigate whether LLMs implicitly encode linear spatial world models, which we define as linear representations of physical space and object configurations. We introduce a formal framework for spatial world models and assess whether such structure emerges in contextual embeddings. Using a synthetic dataset of object positions, we train probes to decode object positions and evaluate geometric consistency of the underlying space. We further conduct causal interventions to test whether these spatial representations are functionally used by the model. Our results provide empirical evidence that LLMs encode linear spatial world models. |
| title | Linear Spatial World Models Emerge in Large Language Models |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2506.02996 |