Research map / Multi-agent systems and games / Cooperative multi-agent RL
STAIRS-Former: Spatio-Temporal Attention with Interleaved Recursive Structure TransFormer for Offline Mulit-task Multi-agent Reinforcement Learning
Jiwon Jeon, Myungsik Cho, Youngchul Sung
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 |
|---|---|---|---|
| add spatial recursive module (ablation) | STAIRS-Former · SMAC seen tasks | win rate: 72.4 -> 77.9 | Improved |
| add temporal module (ablation) | STAIRS-Former · SMAC seen tasks | win rate: 76.2 -> 77.9 | Improved |
| add token dropout (ablation) | STAIRS-Former · SMAC seen tasks | win rate: 76.0 -> 77.9 | Improved |
| add temporal module (ablation) | STAIRS-Former · SMAC unseen tasks | win rate: 60.6 -> 64.0 | Improved |
| add spatial recursive module (ablation) | STAIRS-Former · SMAC unseen tasks | win rate: 60.2 -> 64.0 | Improved |
| add token dropout (ablation) | STAIRS-Former · SMAC unseen tasks | win rate: 61.8 -> 64.0 | Improved |
| add spatial, temporal and dropout together (vs. w/o STD) (ablation) | STAIRS-Former · SMAC total mean | win rate: 57.3 -> 67.4 | Improved |
| add Temporal Focus Layer (ablation) | STAIRS-Former · Marine-Hard dormant neurons | dormant neuron ratio | Improved |
| replace HiSSD with STAIRS-Former | offline MT-MARL · SMAC-v2 total mean | win rate: 24.4 -> 30.3 | Improved |
| replace HiSSD with STAIRS-Former | offline MT-MARL · Marine-Easy Medium-Replay | win rate | Hurt |
Closest papers by meaning
- Recurrent Action Transformer with Memory (ICLR 2026 · similarity 0.87)
- Learning Generalizable Skills from Offline Multi-Task Data for Multi-Agent Cooperation (ICLR 2025 · similarity 0.86)
- STAIR: Addressing Stage Misalignment through Temporal-Aligned Preference Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- AMAGO-2: Breaking the Multi-Task Barrier in Meta-Reinforcement Learning with Transformers (NeurIPS 2024 · similarity 0.86)
- Trajectory-Class-Aware Multi-Agent Reinforcement Learning (ICLR 2025 · similarity 0.86)
- Robust Policy Learning via Offline Skill Diffusion (AAAI 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.
