Signals 4 · free daily AI digest

Nonequilibrium Phases of Repulsive Self-Attention: Chaos, Attention Condensation, and Emergent Locality

Paper recorded by Signals 4 on 2026-09-23 in cs.LG. Abstract reproduced from arXiv; link to the original below.

Published 2026-09-23 on arXiv · recorded by Signals 4 on 2026-09-24

Category: cs.LG · 机器学习 · first seen 2026-09-24

Abstract

We study the nonequilibrium dynamics of a minimal recurrent transformer with $N$ normalized tokens, $Q=K=I$, and a negative value map $V=-I$. Similarity-based attention selects nearby representations, while the negative value map drives tokens away from the selected field. This feedback can continually reorganize both the representation geometry and the attention network. For $d=2$, the tokens lie

Read on arXiv →

#4 most recent of 278 cs.LG papers we have recorded · ↑ newer: Even Sharper Bounds for Transductive Learning and Its Applications · ↓ older: Minimal-Norm Univariate Two-Layer ReLU Classification: Exact Solutions
Cite this page: Nonequilibrium Phases of Repulsive Self-Attention: Chaos, Attention Condensation, and Emergent Locality: the #4 most recent of 278 cs.LG papers we have recorded (as of 2026-09-23). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/nonequilibrium-phases-of-repulsive-self-attention-chaos-attention-condensation-a.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable: papers.json
Related: More cs.LG papers · arXiv signals · All papers · Today in AI
Get 4 AI signals a day by email — free.
Subscribe free → See all plans →
Get 4 AI signals a day by email — free
All models · All repos · By company · Daily editions