ACADEMICarXiv · 2025

CADFusion

CADFusion augments text-to-CAD generation with an iterative visual feedback loop. After each generation, the rendered output is evaluated against the prompt using a VLM, and the model refines the code. Shows significant improvement on complex multi-feature parts.

Input
Text + visual feedback
Output
CadQuery
Venue
arXiv
Year
2025
CAD Arena Status
Scheduled for evaluation on the full 200-prompt benchmark. Results will appear on the leaderboard at launch.