AI Paper Atlas

Research map / Large language models / Agents, code and math

Seek-CAD: A Self-refined Generative Modeling for 3D Parametric CAD Using Local Inference via DeepSeek

Xueyang Li, Jiahao Li, Yu Song, Yunzhong Lou, Xiangdong Zhou

ICLR 2026 · Large language models › Agents, code and math · approach: LLMs for optimization and heuristics

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 changedWhereMeasuredEffect
use Seek-CAD step-wise visual feedback (SVF) instead of 3D-PreMise or CADCodeVerify refinement (ablation)Seek-CAD · 500 test CAD modelsIoGT: 0.6562 -> 0.7226 (vs CADCodeVerify)Improved
compare training-free Seek-CAD with fine-tuned CAD-LlamaSeek-CAD · 500 test CAD modelsNovel / IoGT: IoGT 0.7023 -> 0.7226; Novel 77.64% -> 64.04%Mixed
add one round of SVF refinement (ablation)Seek-CAD · 500 test CAD modelsIoGT: 0.6183 -> 0.7226Improved
add a second SVF round (ablation)Seek-CAD · 500 test CAD modelsPass@2: 0.72 -> 0.55Hurt
add the local CAD corpus (RAG) (ablation)Seek-CAD · 200 CAD modelscompilable output: model A fails to compile entirelyImproved
add knowledge constraint system prompt (ablation)Seek-CAD · 200 CAD modelsPass@1: 0.44 -> 0.68Improved
add vector search and text search in corpus retrieval (ablation)Seek-CAD · 200 CAD modelsIoGT: 0.6464 -> 0.7451 (w/o text search vs full)Improved
add intermediate images in SVF (ablation)Seek-CAD · 200 CAD modelsIoGT: 0.6713 -> 0.7451Improved
add CoT in SVF (ablation)Seek-CAD · 200 CAD modelsG-Score: 3.6120 -> 3.8409Improved
add the ultimate image in SVF (ablation)Seek-CAD · 200 CAD modelsIoGT: 0.7036 -> 0.7451Improved
increase CAD command sequence length (complexity)Seek-CAD · Low vs High lengthIoGT / Novel: IoGT 0.7356 -> 0.6759; Novel 56.25% -> 69.34%Mixed
replace local corpus with DeepCAD dataSeek-CAD vs baselines · DeepCAD (300 test)IoGT: 0.7531 -> 0.8095 (vs CADCodeVerify)Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy