04:00
2026-10-12
machinebrief.com
artificial-intelligence
Cross-Behavior Policy Optimization: Unified Learning from Offline Demonstrations and On- and Off-Policy Rollouts via a Shared Regression Target
A new arXiv paper (2610.12830v1) proposes Cross-Behavior Policy Optimization (CBPO), a method that unifies offline demonstrations and on- and off-policy rollouts under a single regression loss with a …