Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Equivariant Networks for Zero-Shot Coordination
Darius Muglich, Christian Schroeder de Witt, Elise van der Pol, 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 |
|---|---|---|---|
| train with C5 equivariance instead of S5 data augmentation (OP) (ablation) | OP agents · Hanabi XP | average score: 15.32 -> 15.45 | Improved |
| train with D10 equivariance instead of S5 data augmentation (OP) (ablation) | OP agents · Hanabi XP | average score: 15.32 -> 16.08 | Improved |
| use D10 instead of C5 (larger group) (ablation) | equivariant agents · Hanabi XP | average score: 15.45 -> 16.08 | Improved |
| use D10 instead of OP (ablation) | equivariant agents · Hanabi XP | bombout rate: 30.8% -> 19.7% | Improved |
| symmetrize SAD agents at test time with D10 | SAD · Hanabi XP | average score: 2.52 -> 3.61 | Improved |
| symmetrize IQL agents at test time with D10 | IQL · Hanabi XP | average score: 10.53 -> 13.62 | Improved |
| symmetrize OP agents at test time with D10 | OP · Hanabi XP | average score: 15.32 -> 16.48 | Improved |
| symmetrize OBL-L5 agents with C5 | OBL-L5 · Hanabi XP | average score: 23.77 -> 23.77 | No clear effect |
| symmetrize OBL-L5 agents with D10 | OBL-L5 · Hanabi XP | average score: 23.77 -> 23.89 | Improved |
| symmetrize SAD agents with D10 | SAD · Hanabi XP | bombout rate: 87.2% -> 82.0% | Improved |
| symmetrize IQL agents with D10 | IQL · Hanabi XP | bombout rate: 39.4% -> 33.3% | Improved |
| symmetrize OP agents with D10 | OP · Hanabi XP | bombout rate: 30.8% -> 24.6% | Improved |
Closest papers by meaning
- High Entropy Regularization Leads to Symmetry Equivariant Policies in Dec-POMDPs (NeurIPS 2026 · similarity 0.90)
- ${\rm E}(3)$-Equivariant Actor-Critic Methods for Cooperative Multi-Agent Reinforcement Learning (ICML 2024 · similarity 0.87)
- OvercookedV2: Rethinking Overcooked for Zero-Shot Coordination (ICLR 2025 · similarity 0.87)
- Hierarchical Equivariant Policy via Frame Transfer (ICML 2025 · similarity 0.86)
- Boosting Sample Efficiency and Generalization in Multi-agent Reinforcement Learning via Equivariance (NeurIPS 2024 · similarity 0.86)
- Cross-environment Cooperation Enables Zero-shot Multi-agent Coordination (ICML 2025 · similarity 0.85)
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.
