NVIDIA Interns Release SpatialClaw: Code as Action Interface for VLM Spatial Reasoning
An NVIDIA internship team led by Seokju Cho released the SpatialClaw paper, which uses code as an action interface to make vision-language models more flexible at 3D/4D spatial reasoning.
2026-09-29 ~ 2026-09-29 · 2 related posts
- SpatialClaw: Training-Free Code-as-Action Interface Boosts VLM Agentic Spatial Reasoning — CMHungSteven · 2026-09-29
- SpatialClaw Credits: Work Led by Seokju Cho During NVIDIA Internship — CMHungSteven · 2026-09-29