Text-to-CAD generation on Text2CAD (test)
29.72IoUMemory-Augmented Reinforcement Learning Agent
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Memory-Augmented Reinforcement Learning Agentfeedback=closed-loop geometric-kernel feedback, memory=case library and skill library2026.05 | 29.72 | 0.0788 | 0.1216 | |
| CADCodeVerifytype=agent-style generation with execution checking2026.05 | 19.25 | 0.0827 | 0.2871 | |
| Text2CADtype=supervised-learning-based2026.05 | 11.55 | 0.1098 | 0.3609 | |
| cadrilletype=multimodal reconstruction with online RL2026.05 | 2.3 | 0.1524 | 0.4699 |