AI Paper Atlas

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

GeoPredict: Leveraging Predictive Kinematics and 3D Gaussian Geometry for Precise VLA Manipulation

Jingjing Qian, Boyao Han, Chen Shi, Lei Xiao, Long Yang, Shaoshuai Shi, Li Jiang

CVPR 2026 · Vision-language and multimodal › Robot action and navigation · approach: Memory and subgoal-based manipulation · 14 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 history Track Encoder (ablation)pi0 · RoboCasa Human-50success rate (%): 42.3 -> 44.8Improved
Add Future Track Query with L_track (ablation)GeoPredict · RoboCasa Human-50success rate (%): 44.8 -> 47.2Improved
Add depth supervision on initial Gaussians (ablation)GeoPredict · RoboCasa Human-50success rate (%): 47.2 -> 49.4Improved
Add track-guided Gaussian refinement (ablation)GeoPredict · RoboCasa Human-50success rate (%): 50.5 -> 52.4Improved
Render color in addition to depth (ablation)GeoPredict · RoboCasa Human-50success rate (%): 49.4 -> 49.2No clear effect
Double global Gaussian primitives (N_G 4 to 8) (ablation)GeoPredict · RoboCasa Human-50success rate (%) and training time: 49.4 -> 51.4; 12.0 -> 19.1 h/epochMixed
Use track-guided refinement with N'_G=64 instead of global N_G=8 (ablation)GeoPredict · RoboCasa Human-50success rate (%): 51.4 -> 52.4Improved
Add GeoPredict predictive modules to pi0pi0 · RoboCasa Human-50success rate (%): 42.3 -> 52.4Improved
Add GeoPredict predictive modules to pi0pi0 · LIBERO-Longsuccess rate (%): 87.6 -> 94.0Improved
Add GeoPredict predictive modules to pi0pi0 · LIBERO-Goalsuccess rate (%): 94.2 -> 95.7No clear effect
Add GeoPredict predictive modules to pi0pi0 · real-world geometry generalizationsuccess rate (%): 50.0 -> 95.0Improved
Add GeoPredict predictive modules to pi0pi0 · real-world visual robustnesssuccess rate (%): 35.0 -> 90.0Improved

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