Model-based Offline RL via Robust Value-Aware Model Learning with Implicitly Differentiable Adaptive Weighting

Model-based Offline RL via Robust Value-Aware Model Learning with Implicitly Differentiable Adaptive Weighting explores ROMI enhances offline reinforcement learning with a robust, value-aware model learning approach, offering controllable conservatism and stable model updates, outperforming RAMBO on D4RL and NeoRL datasets.. Commercial viability score: 7/10 in Reinforcement Learning.