Sharp Spectral Thresholds for Logit Fixed Points
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.
Code (0)
등록된 구현이 없습니다.
Tasks
Reinforcement LearningSimilar Papers 제목 키워드 기반
Sharp Recovery Thresholds of Tensor PCA Spectral Algorithms
Many applications seek to recover low-rank approximations of noisy tensor data. We consider several practical and effective matricization strategies which construct specific matrices from such tensors and then apply spec…
Spectral Logit Sculpting: Adaptive Low-Rank Logit Transformation for Controlled Text Generation
Entropy-based inference methods have gained traction for improving the reliability of Large Language Models (LLMs). However, many existing approaches, such as entropy minimization techniques, suffer from high computation…
Text GenerationAn Analysis of Logit Learning with the r-Lambert Function
The well-known replicator equation in evolutionary game theory describes how population-level behaviors change over time when individuals make decisions using simple imitation learning rules. In this paper, we study evol…
Imitation LearningIdentification and Estimation of Average Causal Effects in Fixed Effects Logit Models
This paper studies identification and estimation of average causal effects, such as average marginal or treatment effects, in fixed effects logit models with short panels. Relating the identified set of these effects to …
Human Echolocation in Static Situations: Auditory Models of Detection Thresholds for Distance, Pitch, Loudness and Timbre
We investigated, by using auditory models, how three perceptual parameters, loudness, pitch and sharpness, determine human echolocation. We used acoustic recordings from two previous studies, both from stationary situati…