Mechbench measures large language models ability on parallel mechanisms
MechBench: A Domain-Specific LLM Benchmark for Parallel Mechanisms
Machine Learning
Summary
Large language models (LLMs) struggle to handle questions about parallel mechanisms, which require understanding complex details like their shapes, joints, and how they move. The authors created MechBench, a big set of questions designed to test LLMs on this special knowledge and task-solving for those machines. They also made MechGPT, a version of an LLM fine-tuned for these tasks. Their tests show that even the best current models find hard questions challenging, confirming MechBench as a useful tool to measure LLM skills in this area.
What this means in practice
- •For mechanical design engineers: Evaluate AI tools for analyzing and designing complex machinery using MechBench to test reasoning about mechanical constraints and geometry.
- •For ai developers for engineering software: Use MechBench to benchmark and improve language models tailored to help automate mechanism analysis and design tasks.
Authors
Zhihao Xia, Tingting Wang, Hui Li, Yongheng Xing, Chunxu Tian
Abstract
Parallel mechanisms pose distinctive challenges to large language models (LLMs), as valid reasoning depends jointly on mechanism topology, loop closure, geometry, joint types, and modeling assumptions. Yet existing LLM benchmarks provide little systematic coverage of such domain-specific knowledge, constraint-sensitive analysis, and constructive design. To fill this gap, we introduce MECHBENCH, a 2,541-question benchmark for evaluating LLMs in parallel mechanisms across professional knowledge, mechanism analysis, and design. MECHBENCH contains standard and challenge tiers spanning six capability categories and nine executable analysis and design tasks. It is constructed through MECHFORGE, a domain-grounded framework that combines expert-organized knowledge, LLM-assisted question generation, programmatic instance generation, two-stage expert review, and task-aware scoring with exact matching, rubric-based evaluation, and program verification. We further develop MECHGPT as a diagnostic domain-adapted model. Experiments show that strong knowledge performance does not directly translate into executable mechanism reasoning: leading closed-service models achieve over 95 on the standard tier but only 47-60 on the challenge tier, while domain adaptation substantially improves MechGPT over its Qwen3-8B baselines but leaves a large gap in constrained analysis and design. These results establish MECHBENCH as a systematic testbed for measuring domain-specific LLM capabilities in mechanism science.