Entity · technique

Holonomy Memory Reinforcement Learning

techniqueactiveprovisionalholonomy-memory-reinforcement-learning-55c36795·1 events·first seen 35h ago

Aliases: Holonomy Memory Reinforcement Learning

Co-occurring entities

Minimal Markovization via Stable Quotients in Holonomy-Cover Decision Processes

More like this (12)

Physics-EnhAnced Reinforcement Learning Hierarchical Reinforcement Learning Reinforcement Learning with Metacognitive Feedback Supervised Memory Training shielded reinforcement learning Active Offline-to-Online Reinforcement Learning Reinforcement Learning for Code Reinforcement Learning Elicits Contextual Learning of Unseen Language Translation Attention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix It What do Reward Models Memorize?Remember When It Matters: Proactive Memory Agent for Long-Horizon Agents Entropy-Regularized Reinforcement Learning

Recent events (1)

4arXiv · cs.LG·35h ago·source ↗

Minimal Markov sufficient statistics for POMDPs via stable quotients in holonomy-cover decision processes

A new arXiv preprint characterizes the minimal memory required to restore the Markov property in a structured class of POMDPs called holonomy-cover decision processes, where hidden state evolves via permutations applied to a hidden mode. The authors construct a 'stable quotient' abstraction and prove that the current observation paired with the stable class forms an exact finite Markov state, with a minimality guarantee under reachability and decision-separation conditions. The framework enables 'Holonomy Memory Reinforcement Learning,' which uses the stable class as memory and applies standard finite-MDP RL after synchronization, achieving perfect paired-order accuracy with only three memory states in experiments.

Agent and Tool Ecosystem Minimal Markovization via Stable Quotients in Holonomy-Cover Decision Processes Holonomy Memory Reinforcement Learning