构建统一几何空间，对比大模型与人脑网络的拓扑结构相似性。
A Unified Geometric Space for Topological Alignment Between Transformer-Based Models and Human Brain Networks
- 将Transformer注意力拓扑映射到人脑内在连接网络，实现无任务、跨模态比较。
- 151个模型显示连续弧形分布：全局语义模型对齐高级脑区，局部细节模型对齐感官区。
- 发现预训练模型对齐度不随图像精度提升，且微调效果有限，适合认知神经科学与AI可解释性研究。
人工神经网络是否以类似人脑的方式组织信息仍不明确。以往的脑-人工智能对齐研究受限于特定输入和任务，难以进行跨模态比较。本文提出一种脑-模型拓扑对齐空间，将Transformer注意力拓扑映射至人脑内在连接网络（ICNs），实现无任务、模态无关的比较。分析了151个基于Transformer的模型，共62,480个注意力头图谱，观察到连续的弧形分布，反映不同对齐程度。优化全局语义的模型与高阶ICNs对齐，而关注局部细节的模型则与感觉运动ICNs对齐。非直观发现包括DINOv2相比前代模型对齐度下降，以及蒸馏DeiT模型出现反常的缩放反转现象；微调和指令微调对对齐影响甚微。对齐分数与ImageNet准确率无显著相关性（r = 0.266, p = 0.156）。本研究为量化比较人工与生物系统组织原则提供了框架。
原文摘要 · Abstract (English)
Whether artificial neural networks organize information comparably to the human brain remains unclear. Prior brain--AI alignment studies are constrained by specific inputs and tasks, limiting cross-modal comparison. Here we introduce a brain--model topological alignment space, mapping Transformer attention topology onto human intrinsic connectivity networks (ICNs) to enable task-free, modality-agnostic comparison. Analyzing 151 Transformer-based models with 62,480 attention head graphs, we observe a continuous arc-shaped distribution reflecting varying alignment. Models optimized for global semantics aligned with higher-order ICNs, while local-detail models aligned with sensory ICNs. Non-intuitive findings include reduced alignment in DINOv2 compared to its predecessors and a counterintuitive scaling inversion in distilled DeiT models, while fine-tuning and instruction tuning had limited effect. Alignment scores showed no significant correlation with ImageNet accuracy (r = 0.266, p = 0.156). This work offers a quantitative framework for comparing the organizational principles of artificial and biological systems.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。