Research map / Multi-agent systems and games / Cooperative multi-agent RL
Variational Automatic Curriculum Learning for Sparse-Reward Cooperative Multi-Agent Problems
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 rejection-sampling exploration (ablation) | VACL task expansion · Hard-Spread n=4 | coverage rate %: 0 -> 97.2±1.4 | Helped |
| add rejection-sampling exploration (ablation) | VACL task expansion · Simple-Spread n=4 | coverage rate %: 94.1±4.9 -> 96.0±1.9 | Helped |
| seed exploration from newly solved tasks instead of active tasks (ablation) | VACL · Simple-Spread | training progress | Helped |
| use SVGD gradient-based exploration instead of uniform noise (ablation) | VACL · Hard-Spread | test performance | Helped |
| use SVGD gradient-based exploration instead of uniform noise (ablation) | VACL · Simple-Spread n=4 and Push-Ball n=2 | coverage | Mixed |
| increase particle number from 400 to at least 1000 (ablation) | VACL · Simple-Spread n=4 | coverage | Helped |
| use entity progression instead of direct policy transfer (ablation) | VACL · Simple-Spread and Push-Ball | coverage | Helped |
| increase agents first before boxes in entity progression (ablation) | VACL · Lock-and-Return | success rate | Helped |
| sample 95% active / 5% solved tasks instead of uniform sampling (ablation) | VACL · Simple-Spread n=4, Push-Ball n=2 | coverage | Helped |
| diversified queue instead of FIFO queue (ablation) | VACL · MPE | performance | No clear effect |
| VACL vs baselines on sparse-reward Simple-Spread n=8 and Push-Ball n=4 | VACL · MPE | coverage | Helped |
Closest papers by meaning
- Automata-Conditioned Cooperative Multi-Agent Reinforcement Learning (ICML 2026 · similarity 0.89)
- PORTAL: Automatic Curricula Generation for Multiagent Reinforcement Learning (AAAI 2024 · similarity 0.89)
- CQM: Curriculum Reinforcement Learning with a Quantized World Model (NeurIPS 2023 · similarity 0.89)
- Learning Progress Driven Multi-Agent Curriculum (ICML 2025 · similarity 0.88)
- Automated curriculum generation through setter-solver interactions (ICLR 2020 · similarity 0.88)
- Curriculum Design for Trajectory-Constrained Agent: Compressing Chain-of-Thought Tokens in LLMs (NeurIPS 2025 · 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.