Research map / Multi-agent systems and games / Cooperative multi-agent RL
UniMM: A Unified Mixture Model Framework for Multi-Agent Simulation
Longzhong Lin, Xuewu Lin, Kechun Xu, Haojian Lu, Lichao Huang, Rong Xiong, Yue Wang
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 |
|---|---|---|---|
| lengthen prediction horizon from 0.5s to 2s (ablation) | anchor-free K=3, open-loop training · WOSAC val subset (150 scenes) | Realism Meta: 0.7071 -> 0.7388 | Improved |
| lengthen prediction horizon from 0.5s to 4s (ablation) | anchor-based K=2048, open-loop training · WOSAC val subset (150 scenes) | Realism Meta: 0.7361 -> 0.7570 | Improved |
| extend horizon beyond 4s to 8s (ablation) | anchor-based K=2048, open-loop training · WOSAC val subset (150 scenes) | Realism Meta: 0.7570 -> 0.7489 | Hurt |
| increase components from 3 to 6 (ablation) | anchor-free, 4s horizon, open-loop · WOSAC val subset (150 scenes) | Realism Meta: 0.7382 -> 0.7409 | Improved |
| increase components from 6 to 16 (ablation) | anchor-free, 4s horizon, open-loop · WOSAC val subset (150 scenes) | Realism Meta: 0.7409 -> 0.7343 | Hurt |
| increase components from 64 to 2048 (ablation) | anchor-based, 4s horizon, open-loop · WOSAC val subset (150 scenes) | Realism Meta: 0.7388 -> 0.7570 | Improved |
| increase components from 3 to 16 (ablation) | anchor-free, 8s horizon · WOMD val open-loop prediction | mAP: 0.3356 -> 0.2037 | Hurt |
| train on closed-loop samples with closed-loop window equal to 4s horizon (ablation) | anchor-free K=6, 4s horizon · WOSAC val subset (150 scenes) | Realism Meta: 0.7409 -> 0.7376 | Hurt |
| train on closed-loop samples with closed-loop window equal to 4s horizon (ablation) | anchor-based K=2048, 4s horizon · WOSAC val subset (150 scenes) | Realism Meta: 0.7570 -> 0.7475 | Hurt |
| train on closed-loop samples with 0.5s window and 0.5s horizon (ablation) | anchor-based K=2048 · WOSAC val subset (150 scenes) | Realism Meta: 0.7361 -> 0.7658 | Improved |
| train on closed-loop samples with 0.5s window and 0.5s horizon (ablation) | anchor-free K=6 · WOSAC val subset (150 scenes) | Realism Meta: 0.6948 -> 0.7404 | Improved |
| shorten closed-loop window to 0.5s while keeping 4s horizon (ablation) | anchor-free K=6 · WOSAC val subset (150 scenes) | Realism Meta: 0.7376 -> 0.7624 | Improved |
Closest papers by meaning
- UniMotion: A Unified Motion Framework for Simulation, Prediction and Planning (NeurIPS 2025 · similarity 0.87)
- Advancing Multi-agent Traffic Simulation via R1-Style Reinforcement Fine-Tuning (ICLR 2026 · similarity 0.86)
- Causal Composition Diffusion Model for Closed-loop Traffic Generation (CVPR 2025 · similarity 0.86)
- BehaviorGPT: Smart Agent Simulation for Autonomous Driving with Next-Patch Prediction (NeurIPS 2024 · similarity 0.86)
- ReSim: Reliable World Simulation for Autonomous Driving (NeurIPS 2025 · similarity 0.85)
- Model-Based Policy Adaptation for Closed-Loop End-to-end Autonomous Driving (NeurIPS 2025 · similarity 0.85)
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.
