Research map / Reinforcement learning / MDPs and planning
Predictable MDP Abstraction for Unsupervised Model-Based RL
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use PMA's predictable latent action space instead of a classic dynamics model | Zero-shot model-based RL · Seven MuJoCo environments | task performance | Helped |
| Use PMA instead of prior unsupervised model-based methods | Ant tasks · Ant | task success | Helped |
| Use PMA instead of classic models for full-length model-based rollouts | SAC with model-based rollouts · MuJoCo tasks | gap between predicted and true returns | Helped |
| Add the disagreement bonus during unsupervised training (ablation) | PMA · Complex MuJoCo environments | task performance | Helped |
| Add the disagreement penalty during MPPI planning (ablation) | PMA · Complex MuJoCo environments | training stability | Helped |
| Remove the disagreement bonus during unsupervised training (ablation) | PMA · HalfCheetah | task performance | No clear effect |
| Remove the disagreement penalty during MPPI planning (ablation) | PMA · HalfCheetah | task performance | No clear effect |
| Train a classic model on PMA's collected data instead of using a separate latent action space (ablation) | PMA · HalfCheetah | MPPI performance | No clear effect |
| Use a separate latent action space rather than only restricting training data (ablation) | PMA · Unstable environments, including Ant | MPPI performance | Helped |
Closest papers by meaning
- Model-based Reinforcement Learning for Parameterized Action Spaces (ICML 2024 · similarity 0.87)
- Towards General-Purpose Model-Free Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Parallelizing Model-based Reinforcement Learning Over the Sequence Length (NeurIPS 2024 · similarity 0.87)
- Overcoming Slow Decision Frequencies in Continuous Control: Model-Based Sequence Reinforcement Learning for Model-Free Control (ICLR 2025 · similarity 0.87)
- MAMBA: an Effective World Model Approach for Meta-Reinforcement Learning (ICLR 2024 · similarity 0.87)
- The Surprising Ineffectiveness of Pre-Trained Visual Representations for Model-Based Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.