Human-Centered Learning Mechanics: A Dynamical Framework for Entropy-Regulated Representation Learning
The paper introduces Human-Centered Learning Mechanics (HCLM), a framework for entropy-regulated representation learning. It emphasizes the importance of effective entropy in optimizing learning systems under uncertainty and resource constraints. The study presents new insights into the dynamics of information forces and their impact on model training.
- ▪HCLM is proposed as a dynamical and information-theoretic framework for open learning systems.
- ▪The paper formalizes entropy regularization through effective information force and characterizes degenerate entropy regimes.
- ▪Controlled experiments indicate that geometric entropy surrogates can induce stronger and more stable information forces than traditional methods.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Machine Learning arXiv:2605.22940 (cs) [Submitted on 21 May 2026] Title:Human-Centered Learning Mechanics: A Dynamical Framework for Entropy-Regulated Representation Learning Authors:Kim Phuc Tran View a PDF of the paper titled Human-Centered Learning Mechanics: A Dynamical Framework for Entropy-Regulated Representation Learning, by Kim Phuc Tran View PDF HTML (experimental) Abstract:Deep learning is increasingly viewed as a dynamical process in parameter space, yet many existing theories still treat training as a closed optimization system. This view is limited for real-world AI, where models operate under uncertainty, resource constraints, distribution shift, downstream decision risks, and human feedback.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
