arXiv cs.AI· Fengxiao Fan, Jingzhe Ni, Fan Sang, Xiaolong Yin, Yu Liu, Ruofeng Tong, Min Tang, Peng Du·· 11 小时前AI 评分32
TraceCAD：面向智能体 CAD 生成的轨迹引导修复方法
TraceCAD: Trace-Guided Repair for Agentic CAD Generation
AI 导读
TraceCAD 是一个将需求特征、建模步骤、失败证据与候选结果作为持久状态关联的修复层，可诊断可疑操作并在其依赖区域内做有界编辑搜索，再通过执行与保持性校验筛选候选。
正文 · 原文
Abstract:LLM-based CAD agents produce executable parametric programs, but their correction loops may lose evidence about satisfied requirements, faulty operations, and prior repairs. We introduce TraceCAD, a recovery layer that links requested features, modeling steps, failure evidence, and candidate outcomes as persistent state. TraceCAD diagnoses likely faulty operations, searches bounded edits in their dependency regions, validates candidates through execution and preservation checks, and retains successful and failed repair outcomes in reusable skill memory. On DeepCAD-derived benchmarks with 200-model ablations and a 1K-model comparison, TraceCAD achieves competitive geometric quality in terms of IoU, Chamfer distance, and Hausdorff distance. Removing persistent state nearly halves recovery score; removing localized search more than doubles geometric regression and doubles code-agent invocations. Initializing the skill store on disjoint training models further reduces retries, token cost, and latency. These results demonstrate that persistent, localized, and reusable recovery improves final CAD quality and repair reliability.
| Comments: | 17 pages, 7 figures; includes supplementary material; revised manuscript |
| Subjects: | Artificial Intelligence (cs.AI); Graphics (cs.GR); Software Engineering (cs.SE) |
| Cite as: | arXiv:2608.03062 [cs.AI] |
| (or arXiv:2608.03062v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2608.03062 arXiv-issued DOI via DataCite |
Submission history
From: Fengxiao Fan [view email]
[v1]
Tue, 4 Aug 2026 03:23:11 UTC (5,658 KB)
[v2]
Wed, 4 Nov 2026 06:12:15 UTC (5,654 KB)
来源：arXiv cs.AI · arxiv.org