Research map / 3D vision / 3D generation
Dexterous World Models
Byungjun Kim, Taeksoo Kim, Junyoung Lee, Hanbyul Joo
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 real-world fixed-camera data (TASTE-Rob) to TRUMANS training (ablation) | DWM · Real-World Static (TASTE-Rob) | PSNR: 17.959 -> 21.547 | Improved |
| add real-world fixed-camera data to training (ablation) | DWM · Real-World Dynamic (Aria) | LPIPS: 0.543 -> 0.557 | Mixed |
| add real-world data to training (ablation) | DWM · Synthetic Dynamic (TRUMANS) | PSNR: 24.151 -> 25.031 | Improved |
| use rendered hand-mesh video instead of hand mask conditioning (ablation) | DWM · Synthetic Dynamic (hand conditioning ablation) | PSNR: 22.876 -> 24.151 | Improved |
| use hand-mesh video instead of per-frame AdaLN MANO conditioning (ablation) | DWM · Synthetic Dynamic (hand conditioning ablation) | PSNR: 22.789 -> 24.151 | Improved |
| use per-frame AdaLN instead of global AdaLN for hand params (ablation) | DWM · Synthetic Dynamic (hand conditioning ablation) | PSNR: 21.962 -> 22.789 | Improved |
| hand-mesh conditioning vs per-frame AdaLN on LPIPS (ablation) | DWM · Synthetic Dynamic (hand conditioning ablation) | LPIPS: 0.288 -> 0.304 | Hurt |
| DWM vs fine-tuned CogVideoX-Fun without hand-mesh condition | CogVideoX-Fun · Real-World Static (TASTE-Rob) | LPIPS: 0.265 -> 0.227 | Improved |
| DWM vs CVX-Fun Fine-tuned on dynamic real-world SSIM/LPIPS | CogVideoX-Fun · Real-World Dynamic (Aria) | LPIPS: 0.591 -> 0.557 | Mixed |
Closest papers by meaning
- SViMo: Synchronized Diffusion for Video and Motion Generation in Hand-object Interaction Scenarios (NeurIPS 2025 · similarity 0.89)
- MEgoHand: Multimodal Egocentric Hand-Object Interaction Motion Generation (NeurIPS 2025 · similarity 0.89)
- HOIDiffusion: Generating Realistic 3D Hand-Object Interaction Data (CVPR 2024 · similarity 0.87)
- DiffGrasp: Whole-Body Grasping Synthesis Guided by Object Motion Using a Diffusion Model (AAAI 2025 · similarity 0.87)
- ParaHome: Parameterizing Everyday Home Activities Towards 3D Generative Modeling of Human-Object Interactions (CVPR 2025 · similarity 0.87)
- InterAct: Advancing Large-Scale Versatile 3D Human-Object Interaction Generation (CVPR 2025 · similarity 0.87)
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.
