bioRxivNeurosciencePreprint
AI-like parallelogram geometry of word relations in the human neocortex
Abstract
The same semantic relation (e.g., queen-king and woman-man) is encoded as a shared vector direction in language AI, forming parallelogram geometry, whereas its presence in the human brain remains unknown. Our fMRI analysis revealed similar geometry in the neocortex, particularly in higher-order language regions. Our findings link the brain and AI through a common geometry, advancing understanding of how abstract relational knowledge is organized and used in human language processing.
- Citations
- 0
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- A geometric foundation for word meaning in the brain
Hanlin Zhu, Melissa C. Franch, … Sameer Anil Sheth·bioRxiv·Apr 3, 2026
- Neural alignment of knowledge structures relates to human intelligence
Rebekka M. Tenderra, Stephanie Theves·bioRxiv·Apr 13, 2026
- Computational mechanisms for neural representation of words
Ze Fu, Xiaosha Wang, … Yanchao Bi·bioRxiv·Jun 27, 2021
- Causal event structure shapes convergent reactivation signatures in human cortex and language model
Zihan Zhang, YuTing Chen, … Ting Liu·bioRxiv·Oct 7, 2026
- Can computers understand words like humans do? Comparable semantic representation in neural and computer systems
Linmin Zhang, Lingting Wang, … Xing Tian·bioRxiv·Aug 27, 2020
- Distributed Code for Semantic Relations Predicts Neural Similarity
Jeffrey N. Chiang, Yujia Peng, … Martin Max Monti·bioRxiv·Jun 12, 2019