mz516/bert-finetuned-squad-v1

This model is a BERT-based extractive question–answering model fine-tuned on the SQuAD v1.1 dataset. It serves as a general-domain baseline for evaluating performance on standard QA benchmarks and comparing against domain-specific models trained on the MechQA dataset. General-domain baseline model achieving 34.40 EM / 50.25 F1 on MechQA.

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”