Research map / Multi-agent systems and games / Games and equilibria
Multi-Agent Generative Adversarial Imitation 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 centralized MAGAIL instead of behavior cloning | Cooperative navigation · 100–400 expert episodes | average reward: -56.82 -> -46.66; -43.10 -> -23.10; -35.66 -> -21.53; -25.83 -> -15.30 | Helped |
| Use centralized MAGAIL instead of behavior cloning | Cooperative communication · 100–400 expert episodes | average reward: -21.25 -> -15.65; -13.25 -> -7.11; -11.37 -> -7.11; -10.00 -> -7.09 | Helped |
| Use centralized MAGAIL instead of GAIL | Cooperative navigation · 100–400 expert episodes | average reward: -55.01 -> -46.66; -39.21 -> -23.10; -29.89 -> -21.53; -18.76 -> -15.30 | Helped |
| Use centralized MAGAIL instead of GAIL | Cooperative communication · 100–400 expert episodes | average reward: -20.28 -> -15.65; -11.06 -> -7.11; -10.51 -> -7.11; -9.44 -> -7.09 | Helped |
| Use decentralized MAGAIL instead of centralized MAGAIL | Cooperative navigation · 100–400 expert episodes | average reward: -46.66 -> -50.00; -23.10 -> -25.61; -21.53 -> -24.10; -15.30 -> -15.55 | Hurt |
| Use centralized MAGAIL instead of decentralized MAGAIL | Cooperative communication · 100–400 expert episodes | average reward: -18.68 -> -15.65; -8.06 -> -7.11; -8.16 -> -7.11; -7.34 -> -7.09 | Helped |
| Use decentralized MAGAIL instead of GAIL | Predator-Prey, learned agent vs. BC adversary · Predator-Prey | agent reward: -90.55 -> -85.00 | Helped |
| Use decentralized MAGAIL instead of GAIL | Keep-Away, learned agent vs. BC adversary · Keep-Away | agent reward: 26.22 -> 28.73 | Helped |
| Imitate demonstrations with centralized MAGAIL rather than behavior cloning in the changed control environment | Cooperative control · Changed environment, 10 demonstrations | failure rate: 39.8% -> 26.2% | Helped |
| Imitate demonstrations with centralized MAGAIL rather than behavior cloning in the changed control environment | Cooperative control · Changed environment, 10 demonstrations | reward: 1.26 -> 26.57 | Helped |
Closest papers by meaning
- Generative Adversarial Imitation Learning (NeurIPS 2016 · similarity 0.91)
- Mimicking To Dominate: Imitation Learning Strategies for Success in Multiagent Games (NeurIPS 2024 · similarity 0.90)
- Inverse Factorized Soft Q-Learning for Cooperative Multi-agent Imitation Learning (NeurIPS 2024 · similarity 0.89)
- MisoDICE: Multi-Agent Imitation from Mixed-Quality Demonstrations (NeurIPS 2025 · similarity 0.88)
- Hybrid Inverse Reinforcement Learning (ICML 2024 · similarity 0.88)
- Reward-free World Models for Online Imitation Learning (ICML 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.