Research map / Multi-agent systems and games / Cooperative multi-agent RL
Safe Multi-agent Reinforcement Learning with Natural Language Constraints
Ziyan Wang, Meng Fang, Tristan Tomilin, Fei Fang, Yali Du
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 SMALL learned language costs to HAPPO/MAPPO | HAPPO · LaMaSafe-Goal (Ant, easy) | cost: 22.4 -> 5.8 | Improved |
| add SMALL learned language costs to HAPPO/MAPPO | HAPPO · LaMaSafe-Goal (Ant, easy) | reward: 12.5 -> 11.6 | No clear effect |
| add SMALL costs to MAPPO/HAPPO | LaMaSafe-Highway · LaMaSafe-Highway | reward: about 10-15% lower in some cases | Hurt |
| use real constraint embeddings instead of random vectors (ablation) | SMALL-HAPPO · LaMaSafe-Goal (Ant, easy) | cost: 34.5 -> 5.8 | Improved |
| add triplet fine-tuning of the encoder (ablation) | SMALL-HAPPO · LaMaSafe-Goal (Ant, easy) | reward: 6.3 -> 11.6 | Improved |
| add the validation LLM (ablation) | SMALL-HAPPO · LaMaSafe-Goal (Ant, easy) | cost: 10.8 -> 5.8 | Improved |
| add the rule-based descriptor (ablation) | SMALL-HAPPO · LaMaSafe-Goal (Ant, easy) | reward: 7.9 -> 11.6 | Improved |
| add the binary violation flag to the cost (ablation) | SMALL-HAPPO · LaMaSafe-Goal (Ant, easy) | reward: 5.1 -> 11.6 | Improved |
Closest papers by meaning
- From Text to Trajectory: Exploring Complex Constraint Representation and Decomposition in Safe Reinforcement Learning (NeurIPS 2024 · similarity 0.91)
- Learning Agentic Policy from Action Guidance (NeurIPS 2026 · similarity 0.88)
- Enhancing Safety in Reinforcement Learning with Human Feedback via Rectified Policy Optimization (NeurIPS 2025 · similarity 0.88)
- Scalable Constrained Policy Optimization for Safe Multi-agent Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Reinforcing LLM Agents via Policy Optimization with Action Decomposition (NeurIPS 2024 · similarity 0.88)
- Prompted Policy Search: Reinforcement Learning through Linguistic and Numerical Reasoning in LLMs (NeurIPS 2025 · similarity 0.88)
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.
