Research map / Multi-agent systems and games / Human-AI and goal-conditioned
JointDiff: Bridging Continuous and Discrete in Multi-Agent Trajectory Generation
Guillem Capellera, Luis Ferraz, Antonio Rubio Romano, Alexandre Alahi, Antonio Agudo
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 |
|---|---|---|---|
| jointly model possession events with trajectories (vs U2Diff) | multi-agent diffusion · NBA future generation | SADE avg: 2.12 -> 2.01 | Improved |
| jointly model events (vs MoFlow, non-IID) on min SADE | multi-agent diffusion · NFL future generation | SADE min: 2.33 (MoFlow) vs 2.36 | No clear effect |
| joint diffusion vs U2Diff on imputation | multi-agent diffusion · NBA imputation | SADE min: 0.62 -> 0.57 | Improved |
| joint diffusion vs U2Diff on trajectory completion | multi-agent diffusion · Football-U | SADE min: 2.35 -> 1.90 | Improved |
| add joint event modeling (ablation) | JointDiff · NFL text-guided generation | Acc avg: .63 -> .74 | Improved |
| add joint event modeling (ablation) | JointDiff · NFL non-controlled generation | Acc avg: .52 -> .54 | Improved |
| add joint event modeling (ablation) | JointDiff · Bundesliga non-controlled generation | Acc avg: .44 -> .39 | Hurt |
| add joint event modeling (ablation) | JointDiff · NFL text-guided generation | SADE avg: 3.39 -> 3.09 | Improved |
| add weak-possessor guidance via CrossGuid (ablation) | JointDiff · Bundesliga | SADE avg: 3.66 -> 2.85 | Improved |
| use text guidance instead of WPG (ablation) | JointDiff · NFL | SADE avg: 3.26 -> 3.09 | Improved |
| use multinomial instead of absorbing-state discrete diffusion (ablation) | JointDiff · Bundesliga consistency | Acc avg: .70 -> .80 | Improved |
| use multinomial instead of absorbing-state discrete diffusion (ablation) | JointDiff · NBA consistency | Acc max: .99 vs .99 | No clear effect |
Closest papers by meaning
- Sports-Traj: A Unified Trajectory Generation Model for Multi-Agent Movement in Sports (ICLR 2025 · similarity 0.88)
- Event2Tracking: Reconstructing Multi-Agent Soccer Trajectories Using Long-Term Multimodal Context (AAAI 2025 · similarity 0.87)
- Unified Uncertainty-Aware Diffusion for Multi-Agent Trajectory Modeling (CVPR 2025 · similarity 0.86)
- InterControl: Zero-shot Human Interaction Generation by Controlling Every Joint (NeurIPS 2024 · similarity 0.84)
- Generative Trajectory Stitching through Diffusion Composition (NeurIPS 2025 · similarity 0.84)
- LLM-Assisted Semantically Diverse Teammate Generation for Efficient Multi-agent Coordination (ICML 2025 · similarity 0.84)
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.
