Research map / Reinforcement learning / Visual and representation RL
EAT-C: Environment-Adversarial sub-Task Curriculum for Efficient Reinforcement Learning
Shuang Ao, Tianyi Zhou, Jing Jiang, Guodong Long, Xuan Song, Chengqi Zhang
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 |
|---|---|---|---|
| add environment generator (adversarial EG) to the curriculum (ablation) | EAT-C · 2D Pusher, multiple new random environments | Success rate: 80.24 vs 42.23 | Improved |
| add environment generator (adversarial EG) to the curriculum (ablation) | EAT-C · 2D Pusher, training environment | Success rate: 92.04 vs 85.47 | Improved |
| add path-planner (sub-task tree curriculum) to EAT-C (ablation) | EAT-C · 2D Pusher, multiple new random environments | Success rate: 80.24 vs 27.58 | Improved |
| add path-planner (sub-task tree curriculum) to EAT-C (ablation) | EAT-C · 2D Pusher, training environment | Success rate: 92.04 vs 46.02 | Improved |
| EAT-C full method vs. baselines (curriculum of sub-tasks with adversarial EG) | RL agent (SAC) · 7DoF robotic arm | Success rate: 0.873 vs 0.772 (best baseline SGT-PG) | Improved |
| EAT-C curriculum vs. plain SAC | SAC · 2D Pusher, multiple new random environments | Success rate: 80.24 vs 20.83 | Improved |
Closest papers by meaning
- CO-PILOT: COllaborative Planning and reInforcement Learning On sub-Task curriculum (NeurIPS 2021 · similarity 0.89)
- Replay-Guided Adversarial Environment Design (NeurIPS 2021 · similarity 0.89)
- Beyond Fixed Tasks: Unsupervised Environment Design for Task-Level Pairs (AAAI 2026 · similarity 0.89)
- CLUTR: Curriculum Learning via Unsupervised Task Representation Learning (ICML 2023 · similarity 0.89)
- Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning (NeurIPS 2026 · similarity 0.88)
- Adversarial Environment Design via Regret-Guided Diffusion Models (NeurIPS 2024 · similarity 0.87)
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.
