Question Answering Model
Overview
This BERT-based model extracts answers from given context passages in response to questions. Fine-tuned on SQuAD-like datasets, it provides precise span-based answers for reading comprehension tasks.
Model Architecture
Utilizes BERT with 12 layers, 768 hidden units, and 12 attention heads, topped with a question answering head that predicts start and end tokens for answer spans.
Intended Use
Ideal for chatbots, search engines, or educational tools requiring factual extraction from text. It handles English queries and contexts up to 512 tokens.
Limitations
The model may fail on ambiguous questions, out-of-context queries, or non-English text. It assumes the answer is present in the context and could propagate biases from training data.
