AI Paper Atlas

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

NeurIPS 2022 · Reinforcement learning › Policy gradient and actor-critic · approach: Regularized scalable deep value learning · 8 citations (OpenAlex)

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 changedWhereMeasuredEffect
WocaR-DQN vs RADIAL-DQN under PA-ADDQN · Atari BankHeistreward: 508 -> 754Improved
WocaR-DQN vs RADIAL-DQN under PA-ADDQN · Atari RoadRunnerreward: 1290 -> 8239Improved
WocaR-DQN vs RADIAL-DQN under PA-ADDQN · Atari Pongreward: 13.2 -> 19.7Improved
WocaR-DQN vs SA-DQN under PA-ADDQN · Atari Freewayreward: 9.5 -> 12.3Improved
WocaR-DQN vs RADIAL-DQN under PGDDQN · Atari Freewayreward: 33.2 -> 31.4No clear effect
WocaR-DQN vs RADIAL-DQN natural rewardDQN · Atari BankHeistnatural reward: 1178 -> 1220Mixed
WocaR-DQN training time vs SA-DQNDQN · Ataritraining time: 35 -> 18 hoursImproved
WocaR-DQN training time vs RADIAL-DQNDQN · Ataritraining time: 17 -> 18 hoursHurt
WocaR-PPO training steps vs RADIAL-PPO/ATLAPPO · Hopper, Walker2d, Halfcheetahtraining steps: 50% of stepsImproved
WocaR-PPO training steps on AntPPO · Anttraining steps: 75% of stepsImproved
WocaR-PPO training timePPO · Hopper, Walker2d, Halfcheetahtraining time: 50% savedImproved
WocaR-PPO training time on AntPPO · Anttraining time: 32% savedImproved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy