Research map / Reinforcement learning / MDPs and planning
Bayesian Policy Optimization for Model Uncertainty
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 a full posterior belief instead of a point estimate | BPO versus UPMLE · Tiger | average return: 17.9 ± 0.6 vs -9.8 ± 2.0 | Helped |
| use a full posterior belief instead of a point estimate | BPO versus UPMLE · Light-Dark | average return: -166.7 ± 2.4 vs -745.9 ± 22.3 | Helped |
| use a full posterior belief instead of a point estimate | BPO versus UPMLE · Cheetah | average return: 115.6 ± 3.5 vs 108.9 ± 3.3 | Helped |
| use a full posterior belief instead of a point estimate | BPO versus UPMLE · Swimmer | average return: 36.0 ± 0.4 vs 37.6 ± 0.4 | Hurt |
| remove the state and belief encoders (ablation) | BPO · ChainC | average return: 360.1 ± 7.1 -> 231.6 ± 4.2 | Hurt |
| remove the state and belief encoders (ablation) | BPO · Light-Dark | average return: -166.7 ± 2.4 -> -867.9 ± 22.1 | Hurt |
| remove the state and belief encoders (ablation) | BPO · ChainA | average return: 260.1 ± 5.6 -> 268.9 ± 5.7 | No clear effect |
| remove the state and belief encoders (ablation) | BPO · Swimmer | average return: 36.0 ± 0.4 -> 36.9 ± 0.6 | No clear effect |
| use explicit belief reasoning rather than belief-agnostic robust training | BPO versus EPOpt · Tiger | average return: 17.9 ± 0.6 vs -19.9 ± 0.0 | Helped |
| use explicit belief reasoning rather than belief-agnostic robust training | BPO versus EPOpt · Light-Dark | average return: -166.7 ± 2.4 vs -1891.2 ± 45.0 | Helped |
| use a Bayesian policy optimizer rather than a nominal-environment TRPO policy | BPO versus TRPO nominal · Cheetah | average return: 115.6 ± 3.5 vs 64.3 ± 6.1 | Helped |
| increase latent-space discretization to a very fine grid (ablation) | BPO · Chain | average return | Mixed |
Closest papers by meaning
- Bayesian Exploration Networks (ICML 2024 · similarity 0.91)
- Sequential Monte Carlo for Policy Optimization in Continuous POMDPs (NeurIPS 2025 · similarity 0.91)
- Transition Constrained Bayesian Optimization via Markov Decision Processes (NeurIPS 2024 · similarity 0.89)
- Optimizing the Unknown: Black Box Bayesian Optimization with Energy-Based Model and Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Augmented Bayesian Policy Search (ICLR 2024 · similarity 0.88)
- Safeguarded Progress in Reinforcement Learning: Safe Bayesian Exploration for Control Policy Synthesis (AAAI 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.