Entropy-Driven Gradient Stability in Large Language Models: A Non-Equilibrium Thermodynamic Framework for Reinforcement Learning Optimization

Research Square 2026 Preprint Open access

Read full text
Authors: ABDESSAMAD BOURKIBATE
DOI: 10.21203/rs.3.rs-8812364/v1
Cited by: 0

Fields of Study & Subjects

Reinforcement Learning in Robotics Stochastic Gradient Optimization Techniques Topic Modeling

Publication details

Journal / source
Research Square
Publication date
10 February 2026
Language
EN
Document type
Preprint