Research map / Probabilistic and scientific modeling / Molecules and proteins
VectorWorld: Efficient Streaming World Model via Diffusion Flow on Vector Graphs
Chaokang Jiang, Desen Zhou, Jiuming Liu, Li Sun
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 DKAL loss (ablation) | DeltaSim · Waymo NPC policy | ADE: 2.10 -> 1.72 | Improved |
| add DKAL loss (ablation) | DeltaSim · Waymo NPC policy | controllability rho: 0.24 -> 0.53 | Improved |
| add hybrid anchor+residual action head (ablation) | DeltaSim · Waymo NPC policy | ADE: 2.62 -> 1.72 | Improved |
| add hybrid anchor+residual action head (ablation) | DeltaSim · Waymo NPC policy | controllability rho: 0.39 -> 0.53 | Improved |
| add hybrid anchor+residual action head (ablation) | DeltaSim · Waymo NPC policy | latency: 21.7 -> 26.1 | Hurt |
| replace Ctrl-Sim with DeltaSim | NPC policy · Waymo NPC policy | ADE: 2.80 -> 1.72 | Improved |
| replace Ctrl-Sim with DeltaSim | NPC policy · Waymo NPC policy | controllability rho: -0.25 -> 0.53 | Improved |
| replace ScenDream-L with VectorWorld | scene generator · nuPlan scene generation | endpoint distance (m): 0.250 -> 0.078 | Improved |
| replace ScenDream-L with VectorWorld | scene generator · nuPlan scene generation | collision rate (%): 9.30 -> 3.01 | Improved |
| add motion-history warm start (ablation) | VectorWorld closed loop · Waymo closed-loop, IDM | dynamic metric (jerk): 16.6 -> 9.6 | Improved |
| add motion-history warm start (ablation) | VectorWorld closed loop · Waymo closed-loop, IDM | success rate (%): 42.0% -> 78.0% | Improved |
| retrain PPO ego planner inside VectorWorld | PPO · Waymo closed-loop, hardest setting | success rate (%): 25.0% -> 56.0% | Improved |
Closest papers by meaning
- Scenario Dreamer: Vectorized Latent Diffusion for Generating Driving Simulation Environments (CVPR 2025 · similarity 0.88)
- Optimization-Guided Diffusion for Interactive Scene Generation (ECCV 2026 · similarity 0.88)
- ChainFlow-VLA: Causal Flow Planning with Vision-Language Models (NeurIPS 2026 · similarity 0.87)
- AdaWM: Adaptive World Model based Planning for Autonomous Driving (ICLR 2025 · similarity 0.86)
- MAPLE: Latent Multi-Agent Play for End-to-End Autonomous Driving (NeurIPS 2026 · similarity 0.86)
- ReflectDrive-2: Reinforcement-Learning-Aligned Self-Editing for Discrete Diffusion Driving (NeurIPS 2026 · 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.
