Conrad S. Tucker

A large-scale benchmark to assess vision-language model question answering capabilities in engineering simulations

OpenSeeSimE, a large-scale benchmark consisting of 200,000+ question-answer pairs across 10,000 parametrically-varied simulations, is introduced, indicating that deploying VLMs for simulation interpretation will require domain-specific training rather than reliance on general-purpose models, and the benchmark provides...