Research map / Multi-agent systems and games / Cooperative multi-agent RL
VAST: Value Function Factorization with Variable Agent Sub-Teams
Thomy Phan, Fabian Ritz, Lenz Belzner, Philipp Altmann, Thomas Gabor, Claudia Linnhoff-Popien
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 |
|---|---|---|---|
| use VAST with QTRAN operator and meta-gradient assignment instead of flat QTRAN/QMIX/IL | VAST(QTRAN, MetaGrad) · Warehouse[16] | completed orders (training curve) | Improved |
| VAST instead of flat QMIX/QTRAN | VAST · Battle[80] | kill count | Improved |
| VAST instead of flat QMIX/QTRAN/IL | VAST · GaussianSqueeze[800] | system performance | Improved |
| VAST(eta=1/2) instead of QTRAN in small instance | VAST · Warehouse[4] | completed orders | Hurt |
| VAST instead of flat VFF in small instance | VAST · GaussianSqueeze[200] | system performance | No clear effect |
| VFF operator QTRAN instead of IL/VDN/QMIX inside VAST (ablation) | VAST · Warehouse[4], Battle[20] | performance | Improved |
| vary VFF operator inside VAST (ablation) | VAST · GaussianSqueeze[200] | system performance | No clear effect |
| add VAST sub-teams to IL (VAST(IL) vs IL) | IL · Warehouse[4], Battle[20], GaussianSqueeze[200] | learning stability | Improved |
| sub-team ratio eta 1/4 instead of 1/2 (ablation) | VAST · Warehouse[16], GaussianSqueeze[800] | performance over training | Improved |
| sub-team ratio eta 1/2 instead of 1/4 (ablation) | VAST · Battle[80] | kill count | Improved |
| MetaGrad assignment instead of Random/Fixed/Spatial (ablation) | VAST(QTRAN, eta=1/4) · Warehouse[16], Battle[80], GaussianSqueeze[800] | performance | Improved |
| Fixed assignment instead of MetaGrad (ablation) | VAST · Battle[80] | kill count | No clear effect |
Closest papers by meaning
- Towards Understanding Cooperative Multi-Agent Q-Learning with Value Factorization (NeurIPS 2021 · similarity 0.89)
- Learning Nearly Decomposable Value Functions Via Communication Minimization (ICLR 2020 · similarity 0.89)
- A Unified Framework for Factorizing Distributional Value Functions for Multi-Agent Reinforcement Learning (JMLR 2023 · similarity 0.88)
- ConcaveQ: Non-monotonic Value Function Factorization via Concave Representations in Deep Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.88)
- Optimistic Value Instructors for Cooperative Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.86)
- Group-in-Group Policy Optimization for LLM Agent Training (NeurIPS 2025 · 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.
