Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Simplified Action Decoder for Deep 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 |
|---|---|---|---|
| Add the greedy action as an additional input during training | IQL · Matrix game | average reward: around 9.5 -> $9.97 \pm 0.02$ | Helped |
| Add a VDN joint Q-function (ablation) | IQL · Hanabi, 2–5 players | average score: 23.97 $\pm$ 0.01 -> 23.96 $\pm$ 0.01; 23.69 $\pm$ 0.01 -> 23.99 $\pm$ 0.01; 22.76 $\pm$ 0.01 -> 23.79 $\pm$ 0.00; 21.29 | Helped |
| Provide teammates the greedy action as an additional input during training | VDN · Hanabi, 2–5 players | average score: 23.96 $\pm$ 0.01 -> 24.01 $\pm$ 0.01; 23.99 $\pm$ 0.01 -> 23.93 $\pm$ 0.01; 23.79 $\pm$ 0.00 -> 23.81 $\pm$ 0.01; 21.80 | Mixed |
| Add the auxiliary card-status prediction task (ablation) | SAD · Hanabi, 2 players | average score: 24.01 $\pm$ 0.01 -> 24.08 $\pm$ 0.01 | Helped |
| Add the auxiliary card-status prediction task (ablation) | SAD · Hanabi, 3–5 players | average score: 23.93 $\pm$ 0.01 -> 23.81 $\pm$ 0.01; 23.81 $\pm$ 0.01 -> 23.47 $\pm$ 0.01; 23.01 $\pm$ 0.01 -> 22.25 $\pm$ 0.01 | Hurt |
Closest papers by meaning
- A Generalist Hanabi Agent (ICLR 2025 · similarity 0.90)
- Ad-Hoc Human-AI Coordination Challenge (ICML 2025 · similarity 0.88)
- Sparks of Cooperative Reasoning: LLMs as Strategic Hanabi Agents (ICML 2026 · similarity 0.88)
- FightLadder: A Benchmark for Competitive Multi-Agent Reinforcement Learning (ICML 2024 · similarity 0.87)
- Enhancing Cooperative Multi-Agent Reinforcement Learning with State Modelling and Adversarial Exploration (ICML 2025 · similarity 0.86)
- Melting Pot Contest: Charting the Future of Generalized Cooperative Intelligence (NeurIPS 2024 · similarity 0.86)
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.