A project for Chinese Extractive Question Answering (QA), focusing on selecting relevant text spans from a given context to answer specific questions. It supports a two-step process for paragraph and span selection or an end-to-end solution using fine-tuned models like BERT.

AI-powered Question Answering app using Gradio and Transformers

An extractive Question Answering system built with Hugging Face Transformers and Streamlit. Fine-tuned a DistilBERT model on the SQuAD dataset and deployed it as an interactive web app for real-time QA from user-provided context.

Cite or shut up. LLM answers without verifiable sources are liabilities. Extractive RAG with line-anchored citations that fails closed when confidence is low. Deterministic, no API keys.