Research map / Reinforcement learning / Policy gradient and actor-critic
Efficient Adversarial Training without Attacking: Worst-Case-Aware Robust Reinforcement Learning
Yongyuan Liang, Yanchao Sun, Ruijie Zheng, Furong Huang
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 |
|---|---|---|---|
| WocaR-DQN vs RADIAL-DQN under PA-AD | DQN · Atari BankHeist | reward: 508 -> 754 | Improved |
| WocaR-DQN vs RADIAL-DQN under PA-AD | DQN · Atari RoadRunner | reward: 1290 -> 8239 | Improved |
| WocaR-DQN vs RADIAL-DQN under PA-AD | DQN · Atari Pong | reward: 13.2 -> 19.7 | Improved |
| WocaR-DQN vs SA-DQN under PA-AD | DQN · Atari Freeway | reward: 9.5 -> 12.3 | Improved |
| WocaR-DQN vs RADIAL-DQN under PGD | DQN · Atari Freeway | reward: 33.2 -> 31.4 | No clear effect |
| WocaR-DQN vs RADIAL-DQN natural reward | DQN · Atari BankHeist | natural reward: 1178 -> 1220 | Mixed |
| WocaR-DQN training time vs SA-DQN | DQN · Atari | training time: 35 -> 18 hours | Improved |
| WocaR-DQN training time vs RADIAL-DQN | DQN · Atari | training time: 17 -> 18 hours | Hurt |
| WocaR-PPO training steps vs RADIAL-PPO/ATLA | PPO · Hopper, Walker2d, Halfcheetah | training steps: 50% of steps | Improved |
| WocaR-PPO training steps on Ant | PPO · Ant | training steps: 75% of steps | Improved |
| WocaR-PPO training time | PPO · Hopper, Walker2d, Halfcheetah | training time: 50% saved | Improved |
| WocaR-PPO training time on Ant | PPO · Ant | training time: 32% saved | Improved |
Closest papers by meaning
- Beyond Worst-case Attacks: Robust RL with Adaptive Defense via Non-dominated Policies (ICLR 2024 · similarity 0.93)
- Breaking the Barrier: Enhanced Utility and Robustness in Smoothed DRL Agents (ICML 2024 · similarity 0.89)
- Adversarial Diffusion for Robust Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Representation-Based Robustness in Goal-Conditioned Reinforcement Learning (AAAI 2024 · similarity 0.89)
- Off-Policy Actor-Critic for Adversarial Observation Robustness: Virtual Alternative Training via Symmetric Policy Evaluation (ICML 2025 · similarity 0.88)
- Towards Optimal Adversarial Robust Q-learning with Bellman Infinity-error (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.
