aiqbench analysis
This research addresses the critical gap between visually plausible CAD assemblies generated by LLMs and their actual physical functionality, introducing FunCAD-Harness to validate mechanisms through simulation feedback. By benchmarking 46 mechanism families, the study demonstrates that current agents often miss subtle mechanical failures invisible to visual inspection, significantly improving functional success rates in complex design tasks.
Can It Actually Work? Synthesizing Functional CAD Mechanisms
arXiv:2610.26861v1 Announce Type: cross Abstract: While LLM agents can generate visually plausible CAD assemblies, these designs often fail to fulfill their intended physical functions. We study functional CAD design, a setting where agents interpret underspecified requests to construct and iteratively revise mechanisms that satisfy functional requirements in physical simulation. Verifying these requirements is nontrivial: different candidate mechanisms may realize the same function through entirely different components, controls, and operating sequences. To bridge this gap, we propose FunCAD-Harness, a framework that translates abstract requirements into candidate-specific physical tests and uses simulation feedback to guide design revisions. To systematically evaluate this capability, we introduce FunCADBench, a benchmark spanning 46 mechanism families with interdependent and multi-step operations. Experiments across frontier LLM agents show that FunCAD-Harness exposes subtle mechanical failures invisible to visual inspection, substantially improving functional success rates over standard simulation-augmented baselines. Ultimately, our work marks a step toward CAD agents capable of translating human intent into working mechanisms.
Title:Can It Actually Work? Synthesizing Functional CAD Mechanisms
Abstract:While LLM agents can generate visually plausible CAD assemblies, these designs often fail to fulfill their intended physical functions. We study functional CAD design, a setting where agents interpret underspecified requests to construct and iteratively revise mechanisms that satisfy functional requirements in physical simulation. Verifying these requirements is nontrivial: different candidate mechanisms may realize the same function through entirely different components, controls, and operating sequences. To bridge this gap, we propose FunCAD-Harness, a framework that translates abstract requirements into candidate-specific physical tests and uses simulation feedback to guide design revisions. To systematically evaluate this capability, we introduce FunCADBench, a benchmark spanning 46 mechanism families with interdependent and multi-step operations. Experiments across frontier LLM agents show that FunCAD-Harness exposes subtle mechanical failures invisible to visual inspection, substantially improving functional success rates over standard simulation-augmented baselines. Ultimately, our work marks a step toward CAD agents capable of translating human intent into working mechanisms.
Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL); Graphics (cs.GR)
arXiv:2610.26861 [cs.CV]
(or arXiv:2610.26861v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2610.26861 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 260 words
Related coverage
KathDB-FAO: Synthesized Query Plans in a Multimodal DBMS
arXiv · Sep 25, 2026
AutoFraming: Synthesizing Dynamic Frames of C Code with LLMs
arXiv · Oct 22, 2026
Wrong Code, Right Structure: Learning Netlist Representations from Imperfect LLM-Generated RTL
arXiv · Aug 3, 2026
VLPSA: Vision-Language-Poisson-Safe Actions for Full-Body Safety of Learned Policies
arXiv · Sep 22, 2026