04:00
2026-08-12
machinebrief.com
machine-learning
IADD-TR: Intervention-Aware Dynamics Decoupling with Targeted Regularization for Model-Based Reinforcement Learning
Researchers propose IADD-TR, a framework that decouples environment dynamics into action-intervention and action-free stages to reduce policy-induced data bias in model-based reinforcement learning. Eā¦