Research map / Multi-agent systems and games / Cooperative multi-agent RL
Learning Roles With Emergent Social Value Orientations
Wenhao Li, Xiangfeng Wang, Bo Jin, Jingyi Lu, Hongyuan Zha B
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 RESVO role emergence (rank constraint plus SVO-conditioned policy) to LToS (ablation) | LToS · 10-player Cleanup | average extrinsic reward: LToS -> LToS+r | Improved |
| add RESVO role emergence (rank constraint plus SVO-conditioned policy) to D3C (ablation) | D3C · 10-player Cleanup | average extrinsic reward: D3C -> D3C+r | Improved |
| set rank constraint k to 2 (optimal) instead of too small or too large (ablation) | RESVO · 10-player Cleanup | average extrinsic reward: k=2 best; k=1 and k=9 much lower | Improved |
| use rank constraint k=3 or 4 near the optimum of 2 (ablation) | RESVO · 10-player Cleanup | average extrinsic reward: k=3 or 4 promising | No clear effect |
| use SVO-based RESVO instead of sanction-based LIO (static versus dynamic division of labor) | LIO · 10-player Cleanup | extrinsic reward: about two times better | Improved |
| use RESVO instead of D3C or LIO in 3-player Escape Room | D3C and LIO · 3-Player Escape Room | samples to converge: 4 times as many samples | Improved |
| use RESVO instead of ROMA role learning alone | ROMA · Iterated Prisoner's Dilemma | mutual cooperation stability | Improved |
| use RESVO instead of LToS in 3-Player Escape Room | LToS · 3-Player Escape Room | steps per episode | Improved |
Closest papers by meaning
- R3DM: Enabling Role Discovery and Diversity Through Dynamics Models in Multi-agent Reinforcement Learning (ICML 2025 · similarity 0.85)
- Attention-Guided Contrastive Role Representations for Multi-agent Reinforcement Learning (ICLR 2024 · similarity 0.84)
- Multi-agent cooperation through learning-aware policy gradients (ICLR 2025 · similarity 0.83)
- Steering the Herd: A Framework for LLM-based Control of Social Learning (ICLR 2026 · similarity 0.82)
- LOPT: Learning Optimal Pigovian Tax in Sequential Social Dilemmas (NeurIPS 2025 · similarity 0.82)
- An Evolutionary Perspective on AI Alignment (Student Abstract) (AAAI 2025 · similarity 0.82)
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.
