← 研踪 PaperTrailarXiv:2604.09560 · v3
Attention Geometry
首次提交 2026-02-12 01:27
版本更新 2026-10-28 04:17
雷达观测 2026-10-30 10:23
摘要
Attention normalizes learned scores into a Markov operator. In the continuum, the local sector becomes diffusion on a learned Riemannian manifold, but attention natively bonds distant points, creating nonlocal structure such that no tensor geometry can describe. The Bakry-Emery calculus, built directly from the transition operator, extends geometry to these potentially stratified spaces, giving a coordinate-free account of what vision and language transformers learn.
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v3 · 2026-10-28 04:17 · 最新观测版本
Attention Geometry
Attention normalizes learned scores into a Markov operator. In the continuum, the local sector becomes diffusion on a learned Riemannian manifold, but attention natively bonds distant points, creating nonlocal structure such that no tensor geometry can describe. The Bakry-Emery calculus, built directly from the transition operator, extends geometry to these potentially stratified spaces, giving a coordinate-free account of what vision and language transformers learn.
查看此版本解读 →查看此版本原文 ↗