Saved in:
Bibliographic Details
Main Author: Wang, Tongxi
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.15651
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917499305459712
author Wang, Tongxi
author_facet Wang, Tongxi
contents Softmax feedback systems are a common mathematical core of entropy-regularized reinforcement learning, logit game dynamics, population choice, and mean-field variational updates. Their central stability question is simple: when does a self-reinforcing softmax system produce a unique and globally predictable outcome? Classical theory gives a conservative answer. By treating softmax as a unit-scale response, it certifies stability only in a strongly randomized regime. We prove that the classical approach misses an entire stable regime and does not identify the point at which the qualitative change truly occurs. For finite-dimensional affine logit systems, the sharp dimension-free Euclidean threshold is $$β\|ΠWΠ\|_{\mathcal T\to\mathcal T}<2,$$ rather than the previously used condition, which certifies stability only while the softmax system remains safely over-regularized. Our theorem fills the previously missing pre-bifurcation regime, extending stability guarantees for affine softmax feedback systems to reward-responsive yet globally predictable systems. It enlarges the certified stability boundary for these systems and identifies where the model genuinely undergoes a phase transition.
format Preprint
id arxiv_https___arxiv_org_abs_2605_15651
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Sharp Spectral Thresholds for Logit Fixed Points
Wang, Tongxi
Machine Learning
Artificial Intelligence
Computer Science and Game Theory
Softmax feedback systems are a common mathematical core of entropy-regularized reinforcement learning, logit game dynamics, population choice, and mean-field variational updates. Their central stability question is simple: when does a self-reinforcing softmax system produce a unique and globally predictable outcome? Classical theory gives a conservative answer. By treating softmax as a unit-scale response, it certifies stability only in a strongly randomized regime. We prove that the classical approach misses an entire stable regime and does not identify the point at which the qualitative change truly occurs. For finite-dimensional affine logit systems, the sharp dimension-free Euclidean threshold is $$β\|ΠWΠ\|_{\mathcal T\to\mathcal T}<2,$$ rather than the previously used condition, which certifies stability only while the softmax system remains safely over-regularized. Our theorem fills the previously missing pre-bifurcation regime, extending stability guarantees for affine softmax feedback systems to reward-responsive yet globally predictable systems. It enlarges the certified stability boundary for these systems and identifies where the model genuinely undergoes a phase transition.
title Sharp Spectral Thresholds for Logit Fixed Points
topic Machine Learning
Artificial Intelligence
Computer Science and Game Theory
url https://arxiv.org/abs/2605.15651