mz516/xlnet-finetuned-MechQA

This model is an XLNet-based extractive question-answering model fine-tuned on the MechQA dataset, designed for retrieving mechanical material property information from textual context. Achieves 78.21 EM / 84.70 F1 on the MechQA validation set.

This model was developed as part of the study: “Automatic Generation of a Mechanical Properties Question-Answering Dataset for Language Model Benchmarking: A Comparative Study of BERT, XLNet, and LLaMA Models”