Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Unified Multi-Agent Trajectory Modeling with Masked Trajectory Diffusion
Songru Yang, Zhenwei Shi, Zhengxia Zou
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 cached diffusion (CDiff) (ablation) | UniMTD · Basketball-U | minADE20: 1.95 -> 1.88 | Improved |
| add progressive removal (PR) (ablation) | UniMTD · Basketball-U | minADE20: 2.12 -> 1.88 | Improved |
| add complex adaptive router (CAR) instead of Top2 router (ablation) | UniMTD · Basketball-U | minADE20: 2.22 -> 1.88 | Improved |
| add UMoE instead of plain UA layers (ablation) | UniMTD · Basketball-U | minADE20: 2.92 -> 1.88 | Improved |
| add unidirectional masking (UA instead of full attention) (ablation) | UniMTD · Basketball-U | minADE20: 4.44 -> 1.88 | Improved |
| reduce experts to Ne = 4 (ablation) | UniMTD · Basketball-U | minADE20: 1.88 -> 2.13 | Hurt |
| increase experts to Ne = 16 (ablation) | UniMTD · Basketball-U | minADE20: 1.88 -> 1.90 | Mixed |
| increase model capacity to 6 UMoEs (ablation) | UniMTD · Basketball-U | minADE20: 1.88 -> 1.85 | Mixed |
| use 15 denoising steps instead of 5 in CDiff (ablation) | UniMTD · Basketball-U | minADE20: 1.88 -> 2.14 | Hurt |
| use 50 denoising steps instead of 5 in CDiff (ablation) | UniMTD · Basketball-U | minADE20: 1.88 -> 2.73 | Hurt |
| use 2 denoising steps instead of 5 in CDiff (ablation) | UniMTD · Basketball-U | minADE20: 1.88 -> 1.99 | Hurt |
| replace naive DDPM with CDiff (ablation) | UniMTD · Basketball-U | minADE20: 2.69 -> 1.88 | Improved |
Closest papers by meaning
- Unified Uncertainty-Aware Diffusion for Multi-Agent Trajectory Modeling (CVPR 2025 · similarity 0.89)
- UniMotion: A Unified Motion Framework for Simulation, Prediction and Planning (NeurIPS 2025 · similarity 0.88)
- SingularTrajectory: Universal Trajectory Predictor Using Diffusion Model (CVPR 2024 · similarity 0.87)
- DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving (CVPR 2025 · similarity 0.87)
- Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous and Instruction-guided Driving (CVPR 2024 · similarity 0.86)
- Copilot4D: Learning Unsupervised World Models for Autonomous Driving via Discrete Diffusion (ICLR 2024 · 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.
