NEET-BioBERT

About Model

The DistilBERT NEET Biology MCQ Classifier (NEET_BioBERT) is an educational research model fine-tuned on the distilbert-base-uncased architecture to classify and select the correct option among four choices (A, B, C, or D) for NEET-style biology multiple-choice questions. Trained using PyTorch and Hugging Face Transformers for a multiple-choice classification task, the model was configured with a learning rate of 5e-5, a batch size of 4, a weight decay of 0.01, and run for 10 epochs. The training utilized the NEET Biology QA Dataset on huggingface, a domain-specific undergraduate medical entrance exam dataset containing 793 questions split into an 80% training and 20% validation distribution. Upon evaluation, the model achieved a final training loss of approximately 0.35 and a validation accuracy of 72.96% (~73%). Designed primarily for educational research, AI-powered biology assistants, and MCQ practice evaluation, this model serves as a baseline for future fine-tuning with larger datasets. However, because it was trained on a small dataset, it is restricted strictly to biology content and lacks support for assertion-reasoning, diagram-based, or case-study paragraph questions. Distributed under the cc-by-nc-sa-4.0 license, this model remains an experimental baseline and is not recommended as a final exam-ready solution without further validation.