How large language models (LLMs) align with the neural representation and computation of human language is a central question in cognitive science. Using representational geometry as a mechanistic lens, we addressed this by tracking entropy, curvature, and fMRI encoding scores throughout Pythia (70M-1B) training. We identified a geometric modularization where layers self-organize into stable low- and high-complexity clusters. The low-complexity module, characterized by reduced entropy and curvature, consistently better predicted human language network activity. This alignment followed heterogeneous spatial-temporal trajectories: rapid and stable in temporal regions (AntTemp, PostTemp), but delayed and dynamic in frontal areas (IFG, IFGorb). Crucially, reduced curvature remained a robust predictor of model-brain alignment even after controlling for training progress, an effect that strengthened with model scale. These results links training-driven geometric reorganization to temporal-frontal functional specialization, suggesting that representational smoothing facilitates neural-like linguistic processing.

翻译：大型语言模型（LLMs）如何与人类语言的神经表征和计算过程对齐，是认知科学的核心问题。我们以表征几何为机制视角，通过追踪Pythia（70M-1B）训练全程的熵、曲率和fMRI编码分数来探究这一问题。我们发现了一种几何模块化现象：各层自组织为稳定的低复杂度与高复杂度集群。低复杂度模块以较低的熵和曲率为特征，始终能更好地预测人类语言网络活动。这种对齐遵循异质的时空轨迹：在颞叶区域（AntTemp, PostTemp）快速且稳定，但在额叶区域（IFG, IFGorb）则延迟且动态变化。关键的是，即使在控制训练进度后，降低的曲率仍然是模型-大脑对齐的稳健预测因子，且该效应随模型规模增大而增强。这些结果将训练驱动的几何重组与颞叶-额叶功能特化联系起来，表明表征平滑化促进了类神经的语言处理过程。