Research map / Reinforcement learning / Offline RL
Lightweight Online Reinforcement Learning for Block Decomposition of CAD Models
Xitong Luo, Zhenghan Wu, Yucong Wang, Yi Cai
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add frozen pre-trained feature encoder instead of joint training (ablation) | Online SAC framework · Self-built dataset | Total time (min): 39.17 -> 34.12 | Improved |
| add frozen feature encoder instead of joint training (ablation) | Online SAC framework · Public dataset subset | SVR (%): 86.23 -> 90.14 | Improved |
| add frozen feature encoder instead of joint training (ablation) | Online SAC framework · Public dataset subset | RL time (min): 4.51 -> 2.38 | Improved |
| add frozen encoder (ablation) | Online SAC framework · Self-built dataset | SVR (%): 100.00 -> 100.00 | No clear effect |
| add continuous-to-discrete action projection (ablation) | Online SAC framework · Self-built dataset | SVR (%): 86.39 -> 100.00 | Improved |
| add continuous-to-discrete action projection (ablation) | Online SAC framework · Public dataset subset | QHMR (%): 72.23 -> 94.96 | Improved |
| add action projection (ablation) | Online SAC framework · Self-built dataset | Min_steps: 26.37 -> 14.04 | Improved |
| add action projection (ablation) | Online SAC framework · Self-built dataset | RL time (min): 2.44 -> 2.25 | Hurt |
| pure online RL instead of offline RL | Online SAC framework · Cross-paradigm | SVR (%): 90.01 -> 95.79 | Improved |
| pure online RL instead of offline RL | Online SAC framework · Cross-paradigm | Total delivery time (min): 138.09 -> 36.05 | Improved |
| drop offline pre-training relative to Hybrid-FT | Online SAC framework · Cross-paradigm | SVR (%): 96.10 -> 95.79 | No clear effect |
| drop offline pre-training relative to Hybrid-FT | Online SAC framework · Cross-paradigm | Total delivery time (min): 173.65 -> 36.05 | Improved |
Closest papers by meaning
- From Intent to Execution: Multimodal Chain-of-Thought Reinforcement Learning for Precise CAD Code Generation (AAAI 2026 · similarity 0.87)
- cadrille: Multi-modal CAD Reconstruction with Reinforcement Learning (ICLR 2026 · similarity 0.86)
- ReCAD: Reinforcement Learning Enhanced Parametric CAD Model Generation with Vision-Language Models (AAAI 2026 · similarity 0.86)
- CAD-Coder: Text-to-CAD Generation with Chain-of-Thought and Geometric Reward (NeurIPS 2025 · similarity 0.85)
- Mesh-RFT: Enhancing Mesh Generation via Fine-grained Reinforcement Fine-Tuning (NeurIPS 2025 · similarity 0.85)
- StepCAD: Mesh-to-CAD Code Generation via LLM Policy and Geometry-Guided Search (NeurIPS 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
