Research map / 3D vision / 3D generation
EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric Video
Ryan Hoque, Peide Huang, David J. Yoon, Mouli sivapurapu, Jian Zhang
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 |
|---|---|---|---|
| use encoder-decoder instead of decoder-only architecture | Dec + BC · EgoDex trajectory prediction, 2s | average distance (m), K=1: 0.045 -> 0.044 | Improved |
| use flow matching instead of behavior cloning | EncDec + BC · EgoDex trajectory prediction, 2s | average distance (m), K=10: 0.044 -> 0.038 | Improved |
| use flow matching instead of behavior cloning at a single sample | EncDec + BC · EgoDex trajectory prediction, 2s | average distance (m), K=1: 0.044 -> 0.051 | Hurt |
| increase the number of samples K from 1 to 10 | EncDec + FM · EgoDex trajectory prediction, 2s | average distance (m): 0.051 -> 0.038 | Improved |
| shorten prediction horizon from 2s to 1s (ablation) | Dec + BC · EgoDex trajectory prediction | average distance (m): 0.045 -> 0.031 | Improved |
| lengthen prediction horizon from 2s to 3s (ablation) | Dec + BC · EgoDex trajectory prediction | average distance (m): 0.045 -> 0.053 | Hurt |
| add visual goal image conditioning (ablation) | Dec + BC · EgoDex inverse dynamics | final distance (m): 0.062 -> 0.029 | Improved |
| increase model size from 200M to 500M parameters (ablation) | Dec + BC · EgoDex trajectory prediction | average distance (m): 0.045 -> 0.045 | No clear effect |
| increase training dataset size (ablation) | Dec + BC · EgoDex trajectory prediction | average and final distance | Improved |
| evaluate on out-of-distribution task Stamp Paper | Dec + BC · EgoDex OOD tasks | average distance (m): 0.045 -> 0.099 | Hurt |
| lengthen prediction horizon from 1s to 3s (ablation) | EncDec + FM · EgoDex trajectory prediction, K=10 | average distance (m): 0.026 -> 0.047 | Hurt |
Closest papers by meaning
- HandEdit: A Unified Benchmark for Egocentric Human-to-Robot Dexterous Hand Image Editing (NeurIPS 2026 · similarity 0.90)
- Generating 6DoF Object Manipulation Trajectories from Action Description in Egocentric Vision (CVPR 2025 · similarity 0.89)
- Modeling Fine-Grained Hand-Object Dynamics for Egocentric Video Representation Learning (ICLR 2025 · similarity 0.87)
- EgoTac: In-the-wild Tactile Prediction from Egocentric Vision (NeurIPS 2026 · similarity 0.87)
- ManipTrans: Efficient Dexterous Bimanual Manipulation Transfer via Residual Learning (CVPR 2025 · similarity 0.87)
- DexFlyWheel: A Scalable and Self-improving Data Generation Framework for Dexterous Manipulation (NeurIPS 2025 · 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.
