Research map / Multi-agent systems and games / Human-AI and goal-conditioned
SyncMos: Scalable Motion Synchronisation for Multi-Agent Scene Interaction
Lingxiao Li, Dongwon Kim, Lingyan Ruan, Bin Chen, Taesoo Kwon, Taehyun Rhee
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 |
|---|---|---|---|
| use dependency-aware story planner instead of Event-Driven Storytelling baseline | LLM event planner (Qwen-3-235B) · Synchronisation test set | Dependency accuracy (%): 68.4 -> 89.9 | Improved |
| use dependency-aware story planner instead of baseline | LLM event planner (GPT-4o) · Dependency test set | Scenario pass rate (%): 0.0 -> 80.0 | Improved |
| use dependency-aware story planner instead of baseline | LLM event planner (GPT-4o) · Dependency test set | Dependency accuracy (%): 20.5 -> 96.9 | Improved |
| use dependency-aware planner with small or mini backbone | GPT-4o-mini planner · Synchronisation test set | Dependency accuracy (%): 71.4 -> 67.2 | Hurt |
| use dependency-aware planner with Qwen-3-8B | Qwen-3-8B planner · Synchronisation test set | Scenario pass rate (%): 26.7 -> 26.7 | No clear effect |
| use proposed planner when number of events grows | planner token usage · Dependency scenarios | Token usage | Improved |
| add time-warping plus diffusion posterior sampling refinement to LINGO | LINGO · Grasp timing control (+-0.5 to +-1.5 s) | Success rate (%): 0.0 -> 84.7 (+0.5 s) | Improved |
| increase negative timing offset to -1.5 s (ablation) | SyncMos synchronisation module · Grasp timing control | Success rate (%): 88.0 (-0.5 s) -> 37.3 (-1.5 s) | Hurt |
| increase number of agents in handover chain from 2 to 10 (ablation) | SyncMos full pipeline · Chained handover, House and Restaurant scenes | Temporal synchronisation error | No clear effect |
Closest papers by meaning
- Sketchtopia: A Dataset and Foundational Agents for Benchmarking Asynchronous Multimodal Communication with Iconic Feedback (CVPR 2025 · similarity 0.86)
- StoryBox: Collaborative Multi-Agent Simulation for Hybrid Bottom-Up Long-Form Story Generation Using Large Language Models (AAAI 2026 · similarity 0.86)
- Motion-Agent: A Conversational Framework for Human Motion Generation with LLMs (ICLR 2025 · similarity 0.86)
- ESCA: Contextualizing Embodied Agents via Scene-Graph Generation (NeurIPS 2025 · similarity 0.85)
- WorldMemArena: Evaluating Multimodal Agent Memory Through Action–World Interaction (NeurIPS 2026 · similarity 0.85)
- HSI-GPT: A General-Purpose Large Scene-Motion-Language Model for Human Scene Interaction (CVPR 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.
