technique
PAC-Bayes
techniqueactiveprovisional
pac-bayes-87f1ff67·1 events·first seen 16d agoAliases: PAC-Bayes
Co-occurring entities
More like this (12)
Bayesian Optimizationposterior predictive variance minimizationOpenAI PBCPACTLog Probability Bias AnalysisPubMedBERTBayesian Inference and Decision Audits for Public Archives of Frontier AI EvaluationsExact Posterior ScoreMBPPBayesian Optimal Experimental DesignBayesian Multiobjective OptimizationBERTopic
Recent events (1)
ReuseRL: Skill Reuse as Compression in Agentic RL via MDL Principle
ReuseRL formalizes agentic reinforcement learning through the Minimum Description Length (MDL) principle, extracting a shared skill dictionary from successful trajectories and augmenting the RL objective with a segmentation cost that penalizes idiosyncratic, non-reusable behaviors. The authors prove a PAC-Bayes generalization bound for this compression penalty. Evaluated on ALFWorld, TextWorld-Cooking, and Countdown-Stepwise, ReuseRL outperforms vanilla GRPO and round-length baselines on both in-distribution and out-of-distribution tasks.