Research map / Multi-agent systems and games / Cooperative multi-agent RL
Deep Decentralized Multi-task Multi-Agent Reinforcement Learning under Partial Observability
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 hysteretic learning rates to Dec-DRQN (ablation) | Dec-DRQN · MAST 8×8, 2 agents, Pf=0.0 | return: 0.05±0.16 -> 0.76±0.11 | Helped |
| Use hysteresis instead of non-hysteretic learning (ablation) | Dec-DRQN in MAMT · MAMT 5×5, 2 agents, Pf=0.3 | coordination stability | Helped |
| Use hysteresis instead of non-hysteretic learning (ablation) | Dec-DRQN in MAMT · MAMT 6×6 and 7×7, 2 agents, Pf=0.3 | learning success | Helped |
| Use hysteresis instead of non-hysteretic learning (ablation) | Dec-DRQN in MAMT · MAMT 4×4, 3 agents | coordination | Helped |
| Add inter-agent parameter sharing (ablation) | Dec-DRQN · MAMT, 2 agents, Pf=0.3 | performance | No clear effect |
| Replace DRQN with Double-DQN (ablation) | Dec-DRQN · MAMT, 2 agents, Pf=0.3 | performance | No clear effect |
| Vary the hysteretic negative learning rate β (ablation) | Dec-HDRQN · MAMT 6×6, 2 agents, Pf=0.25 | return: β ∈ [0.1,0.6]; β ∈ [0.4,0.5] | Mixed |
| Train one multi-task policy directly on pooled experiences instead of specializing and distilling policies (ablation) | Dec-HDRQN · MAMT tasks 3×3 through 6×6, 2 agents, Pf=0.3 | execution-time performance: 70K epochs; 100K epochs; 500K | Hurt |
| Use a larger or smaller network for direct multi-task learning (ablation) | Multi-HDRQN · MAMT tasks 3×3 through 6×6, 2 agents, Pf=0.3 | performance | No clear effect |
Closest papers by meaning
- Learning to Act in Decentralized Partially Observable MDPs (ICML 2018 · similarity 0.90)
- Attention-Based Recurrence for Multi-Agent Reinforcement Learning under Stochastic Partial Observability (ICML 2023 · similarity 0.89)
- MA$^2$E: Addressing Partial Observability in Multi-Agent Reinforcement Learning with Masked Auto-Encoder (ICLR 2025 · similarity 0.89)
- Automata-Conditioned Cooperative Multi-Agent Reinforcement Learning (ICML 2026 · similarity 0.89)
- Decentralized and Disentangled Task–Role Representation Learning for Generalizable Offline Multi-Agent Meta Reinforcement Learning (ICML 2026 · similarity 0.88)
- Enhancing Cooperative Multi-Agent Reinforcement Learning with State Modelling and Adversarial Exploration (ICML 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.