SketchAgentBench – evaluate LLMs on 3D spatial reasoning
A benchmark for 3D coding agents. 11 models tackle spatial reasoning through code and visual feedback: building scenes, inspecting rendered views, and refining geometry.
Read full article →A benchmark for 3D coding agents. 11 models tackle spatial reasoning through code and visual feedback: building scenes, inspecting rendered views, and refining geometry.
Read full article →