Research map / Multi-agent systems and games / Cooperative multi-agent RL
TarMAC: Targeted Multi-Agent Communication
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 |
|---|---|---|---|
| add targeted attention to communication (ablation) | TarMAC on SHAPES · SHAPES 50×50, find[red,red,green,blue] | success rate: 82.4 -> 85.8 | Helped |
| add targeted attention to communication (ablation) | TarMAC on SHAPES · SHAPES 50×50, find[red] | success rate: 89.5 -> 89.5 | No clear effect |
| add communication | SHAPES navigation · SHAPES 50×50, find[red,red,green,blue] | success rate: 69.1 -> 82.4 | Helped |
| add a second communication round (ablation) | TarMAC on Traffic Junction · Traffic Junction hard | success rate: 84.6 -> 97.1 | Helped |
| add a second communication round (ablation) | TarMAC on Traffic Junction · Traffic Junction easy | success rate: 99.9 -> 99.9 | No clear effect |
| increase beyond 2 communication rounds (ablation) | TarMAC on Traffic Junction · Traffic Junction | success rate | No clear effect |
| use multiple communication rounds rather than only one (ablation) | TarMAC · Traffic Junction hard | success rate | Helped |
| reduce message value size to a single scalar (ablation) | TarMAC · Traffic Junction hard | success rate | No clear effect |
| add targeted attention to communication (ablation) | TarMAC on House3D · House3D find[fireplace] | success rate: 64.3 -> 68.9 | Helped |
| add targeted attention to communication (ablation) | TarMAC on House3D · House3D find[fireplace] | average steps: 101.3 -> 82.5 | Helped |
| replace IC3Net message averaging with TarMAC attention | IC3Net · Predator-Prey, 10 agents, 20×20 | average episode steps: 52.4 -> 41.67 | Helped |
| add a second communication round (ablation) | IC3Net + TarMAC · Predator-Prey, 10 agents, 20×20 | average episode steps: 41.67 -> 35.57 | Helped |
Closest papers by meaning
- Efficient Communication in Multi-Agent Reinforcement Learning with Implicit Consensus Generation (AAAI 2025 · similarity 0.87)
- Multi-Agent Reinforcement Learning with Communication-Constrained Priors (NeurIPS 2025 · similarity 0.87)
- Learning Multi-Agent Communication from Graph Modeling Perspective (ICLR 2024 · similarity 0.86)
- T2MAC: Targeted and Trusted Multi-Agent Communication through Selective Engagement and Evidence-Driven Integration (AAAI 2024 · similarity 0.86)
- Language Grounded Multi-agent Reinforcement Learning with Human-interpretable Communication (NeurIPS 2024 · similarity 0.85)
- PMAC: Personalized Multi-Agent Communication (AAAI 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.