Research map / Multi-agent systems and games / Cooperative multi-agent RL
SMIX(λ): Enhancing Centralized Value Functions for Cooperative Multi-Agent Reinforcement 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 IS-free off-policy SARSA(lambda) lambda-return CVF estimation instead of one-step Q-learning | QMIX · SMAC 3s5z | test win rate (%): 16 -> 61 | Helped |
| use lambda-return CVF estimation instead of one-step | QMIX · SMAC 3s_vs_3z | test win rate (%): 15 -> 84 | Helped |
| use lambda-return CVF estimation instead of one-step | QMIX · SMAC 3m | test win rate (%): 95 -> 99 | Helped |
| replace VDN CVF estimation with SMIX(λ) procedure (ablation) | VDN · SMAC 3s5z | test win rate (%): 1 -> 29 | Helped |
| replace VDN CVF estimation with SMIX(λ) procedure (ablation) | VDN · SMAC 2s3z | test win rate (%): 64 -> 78 | Helped |
| replace QTRAN CVF estimation with SMIX(λ) procedure (ablation) | QTRAN · SMAC 3m | test win rate (%): 29 -> 84 | Helped |
| replace CVF estimation on scenarios where the original method fails entirely (ablation) | COMA/QTRAN/IQL · SMAC 3s5z | test win rate (%): 0 -> 0 | No clear effect |
| use lambda=0.8 instead of lambda=1 (MC return) or lambda=0 (one-step) (ablation) | SMIX(λ) · SMAC 3s5z and 2s_vs_1sc | test win rate (%) | Helped |
| include off-policy data (buffer 1500) instead of on-policy only (buffer 4) (ablation) | SMIX(λ) · SMAC 3s5z | test win rate (%) | Helped |
| increase replay buffer size from 1500 to 5000 (ablation) | SMIX(λ) · SMAC 3s5z and 2s_vs_1sc | test win rate (%) | Hurt |
Closest papers by meaning
- SMACv2: An Improved Benchmark for Cooperative Multi-Agent Reinforcement Learning (NeurIPS 2023 · similarity 0.91)
- Revisiting Some Common Practices in Cooperative Multi-Agent Reinforcement Learning (ICML 2022 · similarity 0.90)
- Locality Matters: A Scalable Value Decomposition Approach for Cooperative Multi-Agent Reinforcement Learning (AAAI 2022 · similarity 0.90)
- Optimistic Value Instructors for Cooperative Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.88)
- Intrinsic Action Tendency Consistency for Cooperative Multi-Agent Reinforcement Learning (AAAI 2024 · similarity 0.88)
- Learning Joint Behaviors with Large Variations (AAAI 2025 · similarity 0.87)
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.