Co-first author Evo-Depth: A Lightweight Depth-Enhanced Vision-Language-Action Model
A lightweight VLA model that incorporates depth cues to improve spatial grounding and embodied manipulation.
Publications
Selected work on spatially grounded VLA models, depth estimation, and embodied perception.
Evo-1: Lightweight Vision-Language-Action Model with Preserved Semantic Alignment
A lightweight VLA model that preserves semantic alignment while improving efficiency for embodied tasks.
Focusable Monocular Depth Estimation
A region-aware monocular depth estimation framework that focuses depth modeling on target regions while preserving coherent global geometry.
Evo-0: Vision-Language-Action Model with Implicit Spatial Understanding
A VLA model that improves embodied manipulation through implicit spatial understanding.