Research map / Reinforcement learning / Policy gradient and actor-critic
Why Dedicated Critics: Eliminating Target Drift in Multi-Constraint RL
Yue Yang, Chenghao Huang, Hao Wang
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 |
|---|---|---|---|
| use dedicated per-constraint critics instead of a mixed cost critic | PPO-Lagrangian · SafetyCarGoal1-v0 | episodic reward: 1.12 -> 14.56 | Improved |
| use dedicated per-constraint critics instead of a mixed cost critic | PPO-Lagrangian · SafetyCarGoal1-v0 | episodic cost: 55.34 -> 21.72 | Improved |
| use dedicated critics instead of a mixed critic | PPO-Lagrangian · SafetyCarButton1-v0 | episodic cost: 107.14 -> 51.40 | Improved |
| use dedicated critics instead of a mixed critic | PPO-Lagrangian · SafetyAntVelocity-v1 | episodic reward: 2821.72 -> 3324.67 | Improved |
| use dedicated critics instead of a mixed critic | PPO-Lagrangian · SafetyAntVelocity-v1 | episodic cost: 28.52 -> 13.01 | Improved |
| use dedicated critics instead of a mixed critic | PPO-Lagrangian · SafetyHalfCheetahV-v1 | episodic cost: 45.82 -> 4.14 | Improved |
| use dedicated critics instead of a mixed critic | PPO-Lagrangian · SafetyHalfCheetahV-v1 | episodic reward: 2234.25 -> 3035.76 | Improved |
| use dedicated critics instead of a mixed critic | PPO-Lagrangian · SafetyPointGoal1-v0 | episodic reward: 15.78 -> 13.03 | Hurt |
| use dedicated critics instead of a mixed critic | PPO-Lagrangian · SafetyPointGoal1-v0 | episodic cost: 52.81 -> 23.97 | Improved |
| use dedicated critics instead of a mixed critic | PPO-Lagrangian · SafetyHopperVelocity-v1 | episodic reward: 1238.83 -> 1002.73 | Hurt |
| use dedicated critics instead of a mixed critic | PPO-Lagrangian · SafetyHopperVelocity-v1 | episodic cost: 17.21 -> 14.87 | Improved |
| use dedicated per-constraint critics instead of a mixed cost critic | PPO-Lagrangian · Power system, 5 constraints | constraint violation | Improved |
Closest papers by meaning
- Risk-Averse Constrained Reinforcement Learning with Optimized Certainty Equivalents (NeurIPS 2025 · similarity 0.86)
- Reward Penalties on Augmented States for Solving Richly Constrained RL Effectively (AAAI 2024 · similarity 0.85)
- Safe Reinforcement Learning using Finite-Horizon Gradient-based Estimation (ICML 2024 · similarity 0.85)
- Pessimism Meets Risk: Risk-Sensitive Offline Reinforcement Learning (ICML 2024 · similarity 0.85)
- Safe Reinforcement Learning with Instantaneous Constraints: The Role of Aggressive Exploration (AAAI 2024 · similarity 0.85)
- Off-Policy Primal-Dual Safe Reinforcement Learning (ICLR 2024 · similarity 0.85)
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.
