Research map / Multi-agent systems and games / Cooperative multi-agent RL
Towards Open Ad Hoc Teamwork Using Graph-based Policy Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 GPL-Q instead of GNN | Open ad hoc teamwork learner · LBF testing | return: 2.07±0.13 -> 2.32±0.22 | Helped |
| Use GPL-SPI instead of GNN-AM | Open ad hoc teamwork learner · Wolfpack testing | return: 30.87±0.95 -> 37.61±1.69* | Helped |
| Use GPL-SPI instead of GNN-AM | Open ad hoc teamwork learner · FortAttack testing | return: 8.12±0.74 -> 16.82±1.92* | Helped |
| Add GNN processing for variable team inputs (ablation) | Single-agent value-based RL · LBF testing | return: 1.41±0.14 -> 2.07±0.13 | Helped |
| Add an agent model and concatenate its action probabilities (ablation) | QL · LBF testing | return: 1.41±0.14 -> 1.22±0.29 | Hurt |
| Add an agent model and concatenate its action probabilities (ablation) | GNN · LBF testing | return: 2.07±0.13 -> 1.80±0.11 | Hurt |
| Use a GNN instead of concatenated type vectors (ablation) | Single-agent value-based RL · Wolfpack testing | return: 20.57±1.95 -> 8.88±1.57 | Hurt |
| Add an agent model and concatenate its action probabilities (ablation) | QL · Wolfpack testing | return: 20.57±1.95 -> 14.24±2.65 | Hurt |
| Add an agent model and concatenate its action probabilities (ablation) | GNN · Wolfpack testing | return: 8.88±1.57 -> 30.87±0.95 | Helped |
| Add a joint-action value model and use predicted actions as weights in action-value computation (ablation) | GNN-AM · Wolfpack testing | return: 30.87±0.95 -> 37.61±1.69* | Helped |
| Add an agent model and concatenate its action probabilities (ablation) | QL · FortAttack testing | return: -3.51±0.60 -> -3.51±1.51 | No clear effect |
| Add a GNN instead of concatenated type vectors (ablation) | Single-agent value-based RL · FortAttack testing | return: -3.51±0.60 -> 7.01±1.63 | Helped |
Closest papers by meaning
- Open Ad Hoc Teamwork with Cooperative Game Theory (ICML 2024 · similarity 0.94)
- N-agent Ad Hoc Teamwork (NeurIPS 2024 · similarity 0.91)
- ROTATE: Regret-driven Open-ended Training for Ad Hoc Teamwork (NeurIPS 2026 · similarity 0.91)
- Ad Hoc Teamwork via Offline Goal-Based Decision Transformers (ICML 2025 · similarity 0.91)
- Minimum Coverage Sets for Training Robust Ad Hoc Teamwork Agents (AAAI 2024 · similarity 0.89)
- NestRL: A Nested Training Regime for Mutual Adaptation in Human–AI Teaming (NeurIPS 2026 · similarity 0.86)
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.