ROMI

Code for ICLR 2026 paper <Model-based Offline RL via Robust Value-Aware Model Learning with Implicitly Differentiable Adaptive Weighting>

The implementation is heavily based on OfflineRL-Kit.

• Prepare your environments as required by OfflineRL-Kit.
• Train and evaluate ROMI by:

python run_example/run_romi.py