Research map / Reinforcement learning / Offline RL
COMBO: Conservative Offline Model-Based Policy Optimization
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 COMBO's conservative model-based value penalty instead of MOPO | MOPO · halfcheetah-jump | average return: 4016.6 -> 5308.7 | Helped |
| use COMBO instead of CQL (add model-generated data with penalty) | CQL · sawyer-door-close | success rate (%): 36.7 -> 98.3 | Helped |
| use COMBO instead of MOPO | MOPO · ant-angle | average return: 2530.9 -> 2776.9 | Helped |
| train COMBO in latent space of variational model | LOMPO · sawyer-door expert | success rate (%): 0.0% -> 96.7% | Helped |
| use COMBO instead of LOMPO | LOMPO · walker_walk | normalized score | No clear effect |
| use COMBO instead of CQL | CQL · D4RL medium-expert halfcheetah | normalized score: 62.4 -> 90.0 | Helped |
| use COMBO instead of MOReL on random-data hopper and walker2d | MOReL · D4RL random | normalized score | Hurt |
| use COMBO across dataset types | existing methods · D4RL | normalized score | Helped |
| use COMBO instead of uncertainty-based estimates (Max Var) for model error (ablation) | MOPO / MOReL · halfcheetah-jump, ant-angle | uncertainty accuracy | Helped |
Closest papers by meaning
- OCEAN-MBRL: Offline Conservative Exploration for Model-Based Offline Reinforcement Learning (AAAI 2024 · similarity 0.93)
- SUMO: Search-Based Uncertainty Estimation for Model-Based Offline Reinforcement Learning (AAAI 2025 · similarity 0.92)
- Constrained Latent Action Policies for Model-Based Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.92)
- Optimistic Model Rollouts for Pessimistic Offline Policy Optimization (AAAI 2024 · similarity 0.92)
- The Edge-of-Reach Problem in Offline Model-Based Reinforcement Learning (NeurIPS 2024 · similarity 0.92)
- A Clean Slate for Offline Reinforcement Learning (NeurIPS 2025 · similarity 0.91)
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.