Research map / Multi-agent systems and games / Cooperative multi-agent RL
Learning Multi-agent Behaviors from Distributed and Streaming Demonstrations
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 gradient-based constraint learning instead of greedy (ME-greedy) | MA-BIRDS · Evader-patroller | SR: 0.000 ± 0.000 -> 0.960 ± 0.010 | Helped |
| Distributed learners instead of centralized BICS | MA-BIRDS · Evader-patroller | SR: 0.960 ± 0.010 -> 0.960 ± 0.005 | No clear effect |
| Distributed learning with no access to ground truth reward vs DLM | MA-BIRDS · Evader-patroller | CVR: 0.040 ± 0.010 -> 0.015 ± 0.005 | Hurt |
| Single-loop distributed online method vs double-loop DLM | MA-BIRDS · Evader-patroller | STPOI: 7.482 -> 1.000 | Helped |
| Single-loop method vs FTL | MA-BIRDS · Evader-patroller | STPOI: 7.500 -> 1.000 | Helped |
| Distributed (one demo per learner) vs centralized BICS processing two demos | MA-BIRDS · Evader-patroller | STPOI: 1.157 -> 1.000 | Helped |
| Compare with ME-greedy time to reach 50% SR | MA-BIRDS · Evader-patroller | 50% SR scaled time: > 318.075 -> 1.000 | Helped |
| Distributed, no-reward method vs FTL | MA-BIRDS · Drone motion planning | SR: 0.981 ± 0.018 vs 0.974 ± 0.025 | No clear effect |
| Single-loop vs FTL scaled time per iteration | MA-BIRDS · Drone motion planning | STPOI: 10.481 -> 1.000 | Helped |
| Single-loop vs DLM scaled time per iteration | MA-BIRDS · Drone motion planning | STPOI: 9.658 -> 1.000 | Helped |
| Distributed without ground truth reward vs ME-gradient | MA-BIRDS · Drone motion planning | SR: 0.980 ± 0.018 vs 0.974 ± 0.025 | No clear effect |
Closest papers by meaning
- Distributed Multi-Agent Bandits Over Erdős-Rényi Random Networks (NeurIPS 2025 · similarity 0.87)
- Multi-agent cooperation through learning-aware policy gradients (ICLR 2025 · similarity 0.86)
- Peer Learning: Learning Complex Policies in Groups from Scratch via Action Recommendations (AAAI 2024 · similarity 0.86)
- Multi-Agent Imitation Learning: Value is Easy, Regret is Hard (NeurIPS 2024 · similarity 0.86)
- A Black-box Approach for Non-stationary Multi-agent Reinforcement Learning (ICLR 2024 · similarity 0.85)
- Impact of Decentralized Learning on Player Utilities in Stackelberg Games (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.