AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

DexVLG: Dexterous Vision-Language-Grasp Model at Scale

Jiawei He, Danshi Li, Xinqiang Yu, Zekun Qi, Wenyao Zhang, Jiayi Chen, Zhaoxiang Zhang, Zhizheng Zhang, Li Yi, He Wang

ICCV 2025 · Vision-language and multimodal › Robot action and navigation · approach: Memory and subgoal-based manipulation · 31 citations (OpenAlex)

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 changedWhereMeasuredEffect
add language instruction input (ablation)DexVLG · LVIS-Seensuccess rate: 84.4 -> 87.7Improved
add language instruction input (ablation)DexVLG · Unseensuccess rate: 64.5 -> 79.1Improved
add language instruction input (ablation)DexVLG · SamPart3Dsuccess rate: 70.2 -> 76.3Improved
add color to point cloud (ablation)DexVLG · LVIS-Seensuccess rate: 47.6 -> 75.3Improved
add color to point cloud (ablation)DexVLG · LVIS-Seenpart grasp accuracy: 15.6 -> 39.1Improved
add color to point cloud (ablation)DexVLG · SamPart3Dsuccess rate: 33.6 -> 53.4Improved
scale training data from 10% to 100% (ablation)DexVLG · LVIS-Seensuccess rate: 49.7 -> 87.7Improved
scale training data from 10% to 100% (ablation)DexVLG · Unseenpart grasp accuracy: 7.9 -> 36.6Improved
increase model size from 255M to 1B (ablation)DexVLG · LVIS-Seensuccess rate: 74.8 -> 75.3No clear effect
use 320M model instead of 255M or 1B (ablation)DexVLG · LVIS-Seenpart grasp accuracy: 35.6 -> 41.6Mixed
add CLIP text embedding to small baselineDGN2.0* · LVIS-Seensuccess rate: 70.8 -> 67.8Hurt
add CLIP text embedding to small baselineDGN2.0* · LVIS-Seenpart grasp accuracy: 38.5 -> 40.5Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy