Research map / Reinforcement learning / Policy gradient and actor-critic
RiskZero: Plan More to Risk Less with a Learned Model
Yousef Yassin, Junfeng Wen
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 |
|---|---|---|---|
| optimize trajectory-level risk with look-ahead search (RiskZero AZ) instead of Sampled TQL | Sampled TQL · SBM 60-30, ER 10% density | 0.25-CVaR return: 38.78 -> 56.86 | Improved |
| optimize trajectory-level risk with look-ahead search (RiskZero AZ) instead of Sampled TQL | Sampled TQL · SMIS 60, ER 15% density | 0.25-CVaR return: 10.58 -> 24.94 | Improved |
| use trajectory-level risk instead of naive future-risk optimization (Risk AZ (N) to Risk AZ) (ablation) | AlphaZero-style planner · SBM 60-30, ER 15% density | 0.25-CVaR return: 1.51 -> 57.34 | Improved |
| use trajectory-level risk instead of naive future-risk optimization (Risk MZ (N) to Risk MZ) (ablation) | MuZero-style planner · SBM 30-10, ER 10% density | 0.25-CVaR return: 0.0 -> 7.40 | Improved |
| add look-ahead search with learned model over Sampled TQL | Sampled TQL · Mini Grid 0.25-CVaR | samples to optimal policy: 3x fewer samples | Improved |
| add RiskZero over Sampled TQL | Sampled TQL · Space Invaders (MinAtar variant) 0.25-CVaR | samples to optimal policy: ~3x more samples for Sampled TQL | Improved |
| use Gumbel planning instead of MuZero heuristic search (ablation) | RiskZero · SBM (30,20,15%) | episode risk: n=2 simulations exceed all MuZero variants; MuZero needs n>16 | Improved |
| increase number of samples D used in search (ablation) | RiskZero (Gumbel and MuZero search) · SBM (30,20,15%) | episode risk | Improved |
| train with risk-neutral objective, evaluate under 0.25-CVaR | RiskZero · Mini Grid | 0.25-CVaR episode risk | Hurt |
Closest papers by meaning
- Risk-Conditioned Reinforcement Learning: A Generalized Approach for Adapting to Varying Risk Measures (AAAI 2024 · similarity 0.91)
- Pessimism Meets Risk: Risk-Sensitive Offline Reinforcement Learning (ICML 2024 · similarity 0.89)
- A Reductions Approach to Risk-Sensitive Reinforcement Learning with Optimized Certainty Equivalents (ICML 2025 · similarity 0.88)
- RobustZero: Enhancing MuZero Reinforcement Learning Robustness to State Perturbations (ICML 2025 · similarity 0.88)
- Beyond CVaR: Leveraging Static Spectral Risk Measures for Enhanced Decision-Making in Distributional Reinforcement Learning (ICML 2025 · similarity 0.88)
- Provable Risk-Sensitive Distributional Reinforcement Learning with General Function Approximation (ICML 2024 · similarity 0.88)
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.
