Research map / Reinforcement learning / Visual and representation RL
Discovering General Reinforcement Learning Algorithms with Adversarial Environment Design
Matthew Thomas Jackson, Minqi Jiang, Jack Parker-Holder, Risto Vuorio, Chris Lu, Gregory Farquhar, Shimon Whiteson, Jakob Nicolaus Foerster
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 adversarial environment design (GROOVE) to LPG meta-training | LPG · unseen Grid-Worlds | fraction of tasks below 75% A2C-normalized return: 187% more tasks fail for LPG | Improved |
| add environment design to LPG | LPG · Grid-World higher-scoring half of tasks | normalized return | No clear effect |
| add environment design to LPG | LPG · Atari | tasks with higher mean score: 17 -> 39 | Improved |
| add environment design to LPG | LPG · Atari (IQM and optimality gap) | aggregate performance | Improved |
| use algorithmic regret instead of uniform scoring (domain randomization) (ablation) | GROOVE · Min-Atar | return | Improved |
| use algorithmic regret instead of optimal-policy regret (ablation) | GROOVE · Min-Atar | return | Improved |
| use L1 value loss or positive value loss instead of uniform scoring (ablation) | GROOVE · Min-Atar | return | No clear effect |
| use positive value loss instead of uniform scoring (ablation) | GROOVE · Min-Atar | return | Hurt |
| use A2C or PPO antagonist instead of random or expert agent in AR (ablation) | GROOVE · Min-Atar | return | Improved |
| use A2C antagonist instead of PPO in AR (ablation) | GROOVE · Min-Atar | return | Improved |
| increase number of random training levels (ablation) | LPG · Min-Atar | return | Improved |
| train on handcrafted instead of random levels (same count) (ablation) | LPG · Min-Atar | return | Mixed |
Closest papers by meaning
- Adversarial Environment Design via Regret-Guided Diffusion Models (NeurIPS 2024 · similarity 0.91)
- Improving Regret Approximation for Unsupervised Dynamic Environment Generation (NeurIPS 2025 · similarity 0.89)
- Can Learned Optimization Make Reinforcement Learning Less Difficult? (NeurIPS 2024 · similarity 0.89)
- On Generalization Across Environments In Multi-Objective Reinforcement Learning (ICLR 2025 · similarity 0.88)
- No Regrets: Investigating and Improving Regret Approximations for Curriculum Discovery (NeurIPS 2024 · similarity 0.88)
- Towards General-Purpose Model-Free Reinforcement Learning (ICLR 2025 · 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.
