MAGIC: Multi-Step Advantage-Gated Causal Influence for Multi-agent Reinforcement Learning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911669179908096 |
|---|---|
| author | Yu, Haohan Cong, Jinmiao Wang, Shengzhi Wang, Lu Liu, Chanjuan |
| author_facet | Yu, Haohan Cong, Jinmiao Wang, Shengzhi Wang, Lu Liu, Chanjuan |
| contents | A key challenge in multi-agent reinforcement learning (MARL) lies in designing learning signals that effectively promote coordination among agents. Designing such signals requires estimating how one agent's current action affects its teammates over future interaction steps. To address this, we introduce Multi-step Advantage-Gated Interventional Causal MARL (MAGIC), a framework that estimates multi-step action effects between agents and selectively converts them into intrinsic rewards. MAGIC uses counterfactual action interventions to compare teammate futures under factual and counterfactual branches, and introduces a gate based on advantage to direct exploration toward beneficial behaviors aligned with the task goal. Experiments on Multi-Agent Particle Environments (MPE) and StarCraft micromanagement benchmarks (SMAC and SMACv2) show that MAGIC consistently outperforms leading prior methods, with average relative final performance improvements of 26.9% and 10.1%, respectively. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_01805 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | MAGIC: Multi-Step Advantage-Gated Causal Influence for Multi-agent Reinforcement Learning Yu, Haohan Cong, Jinmiao Wang, Shengzhi Wang, Lu Liu, Chanjuan Multiagent Systems Machine Learning A key challenge in multi-agent reinforcement learning (MARL) lies in designing learning signals that effectively promote coordination among agents. Designing such signals requires estimating how one agent's current action affects its teammates over future interaction steps. To address this, we introduce Multi-step Advantage-Gated Interventional Causal MARL (MAGIC), a framework that estimates multi-step action effects between agents and selectively converts them into intrinsic rewards. MAGIC uses counterfactual action interventions to compare teammate futures under factual and counterfactual branches, and introduces a gate based on advantage to direct exploration toward beneficial behaviors aligned with the task goal. Experiments on Multi-Agent Particle Environments (MPE) and StarCraft micromanagement benchmarks (SMAC and SMACv2) show that MAGIC consistently outperforms leading prior methods, with average relative final performance improvements of 26.9% and 10.1%, respectively. |
| title | MAGIC: Multi-Step Advantage-Gated Causal Influence for Multi-agent Reinforcement Learning |
| topic | Multiagent Systems Machine Learning |
| url | https://arxiv.org/abs/2605.01805 |