Research map / Multi-agent systems and games / Cooperative multi-agent RL
ELIGN: Expectation Alignment as a Multi-Agent Intrinsic Reward
Zixian Ma, Rose E Wang, Li Fei-Fei, Michael S. Bernstein, Ranjay Krishna
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 ELIGN-self intrinsic reward instead of sparse reward only | SAC decentralized · Cooperative navigation 3v0, partial obs | mean test episode reward: 139.07 -> 155.88 | Improved |
| ELIGN-self instead of curiosity-self intrinsic reward | SAC decentralized · Cooperative navigation 3v0, partial obs | mean test episode reward: 133.93 -> 155.88 | Improved |
| Add ELIGN-adv (misalign with adversary expectations) instead of sparse | SAC decentralized · Predator-prey 2v2, partial obs | mean test episode reward: -4.72 -> -2.51 | Improved |
| Add ELIGN-adv instead of sparse | SAC decentralized · Keep-away 2v2, partial obs | mean test episode reward: 4.58 -> 19.46 | Improved |
| Add ELIGN-adv instead of sparse | SAC decentralized · Google football 3v1 with keeper | mean test episode reward: 0.020 -> 0.025 | Improved |
| Add ELIGN-team instead of sparse | SAC decentralized · Physical deception 2v1, partial obs | mean test episode reward: 93.60 -> 101.72 | Improved |
| Add ELIGN-self instead of sparse | SAC decentralized · Physical deception 2v1, partial obs | mean test episode reward: 93.60 -> 69.91 | Hurt |
| Add ELIGN-team instead of sparse | SAC decentralized · Keep-away 2v2, partial obs | mean test episode reward: 4.58 -> 2.96 | Hurt |
| Misalign with adversaries (ELIGN-adv) rather than align with team (ELIGN-team) (ablation) | SAC decentralized · Predator-prey 2v2, partial obs | mean test episode reward: -7.69 -> -2.51 | Improved |
| Scale agents to 6 in heterogeneous navigation; ELIGN-self vs curiosity-self | SAC decentralized · Heterogeneous navigation 6v0 | mean test episode reward: 702.73 -> 646.70 | Hurt |
| Scale to 5 agents; ELIGN-self instead of sparse | SAC decentralized · Cooperative navigation 5v0 | mean test episode reward: 459.92 -> 498.24 | Improved |
| Add ELIGN-self to centralized SAC under partial observability | SAC centralized · Cooperative navigation 3v0 | mean test episode reward: 113.25 -> 137.14 | Improved |
Closest papers by meaning
- Learning to Collaborate with Unknown Agents in the Absence of Reward (AAAI 2025 · similarity 0.87)
- Settling Decentralized Multi-Agent Coordinated Exploration by Novelty Sharing (AAAI 2024 · similarity 0.87)
- Intrinsic Action Tendency Consistency for Cooperative Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.86)
- Multi-agent cooperation through learning-aware policy gradients (ICLR 2025 · similarity 0.86)
- Wonder Wins Ways: Curiosity-Driven Exploration through Multi-Agent Contextual Calibration (NeurIPS 2025 · similarity 0.86)
- Individual Contributions as Intrinsic Exploration Scaffolds for Multi-agent Reinforcement Learning (ICML 2024 · 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.
