Causal event structure shapes convergent reactivation signatures in human cortex and language model
- 1Research Center for Social Computing and Interactive Robotics, Harbin Institute of Technology
Abstract
Humans adapt to changing environments by understanding the latent causal structure of events, and modern large language models (LLMs) increasingly exhibit analogous abilities. However, elucidating how causal knowledge is represented and integrated in the human brain and in LLMs poses a substantial challenge for cognitive science. To address this, we combine topological analysis of representational manifolds with neuroscientific analyses of reactivation. We find that event-level causal relations modulate temporally structured neural reactivation in the human brain, particularly in the angular gyrus, precuneus and prefrontal cortex. By contrast, LLMs express these relations through positionally structured representational reactivation. Despite these differences, the resulting reactivation patterns exhibit similar intrinsic manifold structure across the two systems. Together, these findings suggest that the human brain and LLM share a common principle for organizing causal knowledge.
Related
- Reactivation of neural signatures of food during REM-like sleep in the high-level visual cortex of monkeys
- Reduced causal connectivity in multiple resting-state networks in autism spectrum disorder detected by convergent cross mapping
- Convergent Mitochondrial and Oxidative Stress Signatures in Spaceflight and Tauopathy: A Comparative Multi-Omics Reanalysis in Drosophila melanogaster
- Genomic signatures of positive selection and relaxed selective constraints associated with convergent dietary specialisation
- Sleep reactivation shapes memories of event and its temporal structure
BibTeX
@misc{zhang2026causal,
title = {{Causal event structure shapes convergent reactivation signatures in human cortex and language model}},
author = {Z. Zhang and Y. Chen and Q. Liu and X. Zhao and X. Ding and X. Fei and J. Du and Y. Bao and J. Gao and X. Huang and B. Cai and X. Liang and B. Qin and T. Liu},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.28.754904},
url = {https://www.biorxiv.org/content/10.64898/2026.09.28.754904v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Causal event structure shapes convergent reactivation signatures in human cortex and language model
AU - Z. Zhang
AU - Y. Chen
AU - Q. Liu
AU - X. Zhao
AU - X. Ding
AU - X. Fei
AU - J. Du
AU - Y. Bao
AU - J. Gao
AU - X. Huang
AU - B. Cai
AU - X. Liang
AU - B. Qin
AU - T. Liu
PY - 2026
DA - 2026/09/28
PB - bioRxiv
DO - 10.64898/2026.09.28.754904
UR - https://www.biorxiv.org/content/10.64898/2026.09.28.754904v1
AB - Humans adapt to changing environments by understanding the latent causal structure of events, and modern large language models (LLMs) increasingly exhibit analogous abilities. However, elucidating how causal knowledge is represented and integrated in the human brain and in LLMs poses a substantial challenge for cognitive science. To address this, we combine topological analysis of representational manifolds with neuroscientific analyses of reactivation. We find that event-level causal relations modulate temporally structured neural reactivation in the human brain, particularly in the angular gyrus, precuneus and prefrontal cortex. By contrast, LLMs express these relations through positionally structured representational reactivation. Despite these differences, the resulting reactivation patterns exhibit similar intrinsic manifold structure across the two systems. Together, these findings suggest that the human brain and LLM share a common principle for organizing causal knowledge.
ER -