Research map / Reinforcement learning / Policy gradient and actor-critic
On-Policy Model Errors in Reinforcement Learning
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 on-policy corrections instead of the learned dynamics model alone | MBPO · MuJoCo continuous control | evaluation return | No clear effect |
| Use on-policy corrections instead of the learned dynamics model alone | MBPO · PyBullet continuous control | evaluation return | Helped |
| Use on-policy corrections rather than the learned model alone | MBPOepisodic · CartPole with sine/cosine pole-angle representation | policy learning | Helped |
| Use a one-step rollout horizon instead of longer rollouts (ablation) | On-policy corrections · HalfCheetah | evaluation return | Hurt |
| Increase the number of simulated transitions (ablation) | On-policy corrections · HalfCheetah | asymptotic return | Helped |
Closest papers by meaning
- When to Trust Your Model: Model-Based Policy Optimization (NeurIPS 2019 · similarity 0.92)
- Policy-conditioned Environment Models are More Generalizable (ICML 2024 · similarity 0.90)
- Constrained Latent Action Policies for Model-Based Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Efficient Policy Evaluation with Offline Data Informed Behavior Policy Design (ICML 2024 · similarity 0.89)
- The Edge-of-Reach Problem in Offline Model-Based Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Any-step Dynamics Model Improves Future Predictions for Online and Offline Reinforcement Learning (ICLR 2025 · similarity 0.89)
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.