RoBERTa models domain-adapted with optoelectronics literature; fine-tuned for question-answering, text classification, and text classification.