
The proposed DOMAIN algorithm incorporates an adaptive sampling distribution of model data to achieve mildly conservative value estimation, outperforming prior model-based offline RL methods on benchmark tasks.


coremsg

mildly-conservative-model-based-offline-reinforcement-learning-algorithm-outperforms-prior-methods-on-benchmark-tasks


Mildly Conservative Model-Based Offline Reinforcement Learning Algorithm Outperforms Prior Methods on Benchmark Tasks
