Research map / Multi-agent systems and games / Cooperative multi-agent RL
System Neural Diversity: Measuring Behavioral Heterogeneity in Multi-Agent Learning
Matteo Bettini, Ajay Shankar, Amanda Prorok
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 |
|---|---|---|---|
| increase number of equidistant agents n from 2 to 8 | SND · Goal Navigation | SND: 1.51 -> 1.43 | No clear effect |
| increase number of equidistant agents n from 2 to 8 | HSE · Goal Navigation | HSE: 1.51 -> 5.21 | Hurt |
| increase agents per behavioral cluster from n=2 to n=8 (2 clusters) | SND · Goal Navigation | SND: 1.49 -> 0.81 | Improved |
| increase agents per behavioral cluster from n=2 to n=8 (2 clusters) | HSE · Goal Navigation | HSE: 1.49 -> 1.91 | Hurt |
| constrain SND to 1.33 (optimal) with DiCo versus unconstrained heterogeneous | MAPPO · Differential Steering | reward | Improved |
| constrain SND to too low (0.5) or too high (2) values (ablation) | MAPPO with DiCo · Differential Steering | reward | Hurt |
| add one-hot index observation to homogeneous agents (ablation) | homogeneous MAPPO · Differential Steering | reward | Hurt |
| use heterogeneous instead of homogeneous policies (ablation) | GPPO · Different Size Joint Passage | reward | Improved |
| use heterogeneous instead of homogeneous policies (ablation) | GPPO · Asymmetric Payload Joint Passage | reward under observation noise | Improved |
| use heterogeneous instead of homogeneous policies (ablation) | GPPO · Asymmetric Payload Joint Passage | reward without noise | No clear effect |
| use heterogeneous instead of homogeneous policies after wind reintroduced (ablation) | GPPO · Flocking in Wind | reward | Improved |
| use homogeneous training when SND is about 0 (ablation) | IPPO vs HetIPPO · Goal Navigation (1 goal) | training time | Improved |
Closest papers by meaning
- Controlling Behavioral Diversity in Multi-Agent Reinforcement Learning (ICML 2024 · similarity 0.91)
- HyperMARL: Adaptive Hypernetworks for Multi-Agent RL (NeurIPS 2025 · similarity 0.88)
- Quality Diversity through Human Feedback: Towards Open-Ended Diversity-Driven Optimization (ICML 2024 · similarity 0.88)
- Discovering Creative Behaviors through DUPLEX: Diverse Universal Features for Policy Exploration (NeurIPS 2024 · similarity 0.87)
- Empirical Study on Robustness and Resilience in Cooperative Multi-Agent Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- SEEA-R1: Tree-Structured Reinforcement Fine-Tuning for Self-Evolving Embodied Agents (NeurIPS 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.
