Research map / Reinforcement learning / Visual and representation RL
Environment Generation for Zero-Shot Compositional Reinforcement Learning
Izzeddin Gur, Natasha Jaques, Yingjie Miao, Jongwook Choi, Manoj Tiwari, Honglak Lee, Aleksandra Faust
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 CoDE instead of ALP | web navigation agent · gMiniWoB Login | success rate: 23% -> 92% | Improved |
| use CoDE instead of ALP | web navigation agent · gMiniWoB Address | success rate: 15% -> 98% | Improved |
| use CoDE instead of PAIRED | web navigation agent · gMiniWoB Payment | success rate: 16% -> 93% | Improved |
| use CoDE instead of PAIRED | web navigation agent · gMiniWoB Flight | success rate: 20% -> 95% | Improved |
| use CoDE instead of ALP | web navigation agent · gMiniWoB Shopping | success rate: 6% -> 95% | Improved |
| add the difficulty objective to regret-based generation (ablation) | CoDE on cMiniGrid · cMiniGrid | task success rate | Improved |
| use the difficulty objective alone, without PopRegret (ablation) | CoDE · complex web tasks such as Shopping and Flight | task success rate | Hurt |
| set β=0.2 and α=0.8 (ablation) | CoDE · gMiniWoB Login | success rate: 92% -> 57% | Hurt |
| set β=0.2 and α=0.9 (ablation) | CoDE · gMiniWoB Login | success rate: 92% -> 50% | Hurt |
| use M=4 rollouts and α=0.8 (ablation) | CoDE · gMiniWoB Login | success rate: 92% -> 66% | Hurt |
| use M=4 rollouts and α=0.9 (ablation) | CoDE · gMiniWoB Login | success rate: 92% -> 84% | Hurt |
| use M=6 rollouts and α=0.8 (ablation) | CoDE · gMiniWoB Login | success rate: 92% -> 62% | Hurt |
Closest papers by meaning
- Robust Subtask Learning for Compositional Generalization (ICML 2023 · similarity 0.88)
- Emergent Complexity and Zero-shot Transfer via Unsupervised Environment Design (NeurIPS 2020 · similarity 0.88)
- Don't Just Fine-tune the Agent, Tune the Environment (ICLR 2026 · similarity 0.87)
- Modeling Unseen Environments with Language-guided Composable Causal Components in Reinforcement Learning (ICLR 2025 · similarity 0.87)
- Ground-Compose-Reinforce: Grounding Language in Agentic Behaviours using Limited Data (NeurIPS 2025 · similarity 0.87)
- DRED: Zero-Shot Transfer in Reinforcement Learning via Data-Regularised Environment Design (ICML 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.
