aysin/GORMPO

This repository contains model checkpoints for Generative OOD-regularized Model-based Policy Optimization (GORMPO) on author-created sparse medium-expert level D4RL datasets from the Gym environment: halfcheetah, walker2d and hopper.

Each folder contains density model checkpoints (KDE, VAE, RealNVP, Diffusion, NeuralODE) and policy checkpoints for MBPO-based (mbpo_based) and MOBILE-based (mobile_based) GORMPO. Each policy folder contains policies that are trained with each density model's reward penalty.

Usage

Clone the GORMPO codebase, download the models, and follow the instructions in the README. See the datasets here.

Citation

If you find this repository useful for your research, please cite:

@inproceedings{tumay2026gormpo,
    title={Generative {OOD}-regularized Model-based Policy Optimization},
    author={Tumay, Aysin and Huang, Jiahe and Jortberg, Elise and Yu, Rose},
    booktitle={Advances in Neural Information Processing Systems},
    year={2026},
    url={https://openreview.net/forum?id=3KDoYUU2CU}
}