Research map / Reinforcement learning / Policy gradient and actor-critic
Conservative and Adaptive Penalty for Model-Based Safe 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 |
|---|---|---|---|
| Add an uncertainty-based cost penalty and adapt its scale using environment cost feedback | Model-based safe RL · HalfCheetah | Violation: 202.8 -> 1.7 | Helped |
| Add an uncertainty-based cost penalty and adapt its scale using environment cost feedback | Model-based safe RL · Car-Racing | Violation: 581.0 -> 93.3 | Helped |
| Use an adaptive penalty rather than a fixed κ (ablation) | CAP · HalfCheetah | Violation: 70.0 -> 1.7 | Helped |
| Use an adaptive penalty rather than a fixed κ (ablation) | CAP · Car-Racing | Violation: 149.0 -> 93.3 | Helped |
| Set κ to 10.0 rather than adapting it (ablation) | CAP · HalfCheetah | Return: 1456.3 -> -36.5 | Hurt |
| Set κ to 10.0 rather than adapting it (ablation) | CAP · Car-Racing | Return: 21.7 -> 1.0 | Hurt |
| Set κ to 0.0, removing the uncertainty penalty (ablation) | CAP · HalfCheetah | Violation: 1.7 -> 80.0 | Hurt |
| Set κ to 0.0, removing the uncertainty penalty (ablation) | CAP · Car-Racing | Violation: 93.3 -> 184.0 | Hurt |
| Set κ to 1.0 rather than adapting it (ablation) | CAP · Car-Racing | Violation: 30.3 -> 93.3 | Helped |
Closest papers by meaning
- Sampling-Based Safe Reinforcement Learning (NeurIPS 2026 · similarity 0.92)
- Safe Reinforcement Learning with Instantaneous Constraints: The Role of Aggressive Exploration (AAAI 2024 · similarity 0.91)
- Reward Penalties on Augmented States for Solving Richly Constrained RL Effectively (AAAI 2024 · similarity 0.91)
- Constrained Reinforcement Learning Under Model Mismatch (ICML 2024 · similarity 0.91)
- P2BPO: Permeable Penalty Barrier-Based Policy Optimization for Safe RL (AAAI 2024 · similarity 0.91)
- Off-Policy Primal-Dual Safe Reinforcement Learning (ICLR 2024 · 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.