Research map / Reinforcement learning / Offline RL
One Risk to Rule Them All: A Risk-Sensitive Perspective on Model-Based Offline Reinforcement Learning
Marc Rigter, Bruno Lacerda, Nick Hawes
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| remove risk-averse sampling (ablation) | 1R2R_CVaR · Currency Exchange | static CVaR0.1: 31.4 -> 64.0 | Improved |
| remove risk-averse sampling (ablation) | 1R2R_CVaR · HIV Treatment | static CVaR0.1: 41.2 -> 53.9 | Improved |
| remove risk-averse sampling (ablation) | 1R2R_CVaR · D4RL Medium-Replay Hopper | normalised expected value: 55.3 -> 92.9 | Improved |
| remove risk-averse sampling (ablation) | 1R2R_CVaR · D4RL Random Hopper | normalised expected value: div. -> 30.9 | Improved |
| remove risk-averse sampling (ablation) | 1R2R_CVaR · Stochastic MuJoCo Medium-Replay Hopper (High) | normalised static CVaR0.1: 30.5 -> 51.4 | Improved |
| remove risk-averse sampling (ablation) | 1R2R_CVaR · Currency Exchange | average return: 99.7 vs 78.8 | Hurt |
| remove model ensemble (ablation) | 1R2R_CVaR · D4RL Medium-Expert Hopper | normalised expected value: 45.8 -> 81.6 | Improved |
| remove model ensemble (ablation) | 1R2R_CVaR · D4RL Medium HalfCheetah | normalised expected value: 60.3 -> 74.5 | Improved |
| use Wang risk measure instead of CVaR | 1R2R · Stochastic MuJoCo Medium-Replay Walker2D (High) | normalised static CVaR0.1: 46.0 -> 28.8 | Hurt |
| use larger ensemble of 20 networks (ablation) | 1R2R_CVaR · Stochastic MuJoCo Medium-Replay Walker2D (High) | normalised static CVaR0.1: 46.0 -> 39.1 | Hurt |
| sample 50 successor states instead of fewer (ablation) | 1R2R_CVaR · Stochastic MuJoCo Medium-Replay Walker2D (High) | normalised static CVaR0.1: 46.0 -> 39.6 | Hurt |
| train entirely on synthetic data (f=0) (ablation) | 1R2R_CVaR · Stochastic MuJoCo Medium-Replay Walker2D (High) | normalised static CVaR0.1: 46.0 -> 1.8 | Hurt |
Closest papers by meaning
- Pessimism Meets Risk: Risk-Sensitive Offline Reinforcement Learning (ICML 2024 · similarity 0.92)
- Selective Uncertainty Propagation in Offline RL (AAAI 2025 · similarity 0.91)
- Bayesian Design Principles for Offline-to-Online Reinforcement Learning (ICML 2024 · similarity 0.91)
- A Unified Principle of Pessimism for Offline Reinforcement Learning under Model Mismatch (NeurIPS 2024 · similarity 0.90)
- Information-Directed Offline-to-Online Reinforcement Learning (NeurIPS 2026 · similarity 0.90)
- From Static Policies to Adaptive Priors in Offline Reinforcement Learning (NeurIPS 2026 · similarity 0.90)
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.
