Yuhang's AI Journey
Welcome to Yuhang Wu’s Personal Blog!
I’m Yuhang Wu, a student at the University of Manchester with a strong interest in Retrieval Augmented Generation and Large Language Models. This blog is my space to document my learning journey, sharing insights, study notes, and project experiences as I explore these fascinating areas.
While I’m still in the process of learning and exploring, I believe that each step forward is an important part of growth. I hope that my posts can offer you some inspiration or useful information, and I’m always open to discussions and exchange of ideas.
Thank you for taking the time to visit. I hope you find something interesting and valuable here!
- TableRAG: Enhanced Retrieval-Augmented Generation for Tables with Advanced Filter and Clarifier
- TextRAG: A High-Performance System for Unstructured Data Retrieval
- EasyRAG: A Simplified and Accelerated Framework for RAG Applications
- RAG Augmentation Methods: survey
- RAG End2end Frame and Vector DB Summary
- Dataset-for-Question-Answering
- LLM for table data enhancement
-
TableRAG:
TableRAG is an advanced system designed for handling complex table-based question answering tasks. The system combines cutting-edge large language models (LLMs) with Retrieval-Augmented Generation (RAG) techniques, optimized specifically for multi-table environments. This project addresses the challenges posed by real-world scenarios, where users may need to query a set of related tables rather than a single one, by integrating advanced filtering, clarification, and retrieval mechanisms.
-
TextRAG:
TextRAG is a high-performance retrieval and generation system specifically designed for processing unstructured text data. By integrating key techniques such as Context Segmentation, Context Generation, and Chunks Integration, TextRAG excels in delivering precise and comprehensive answers to complex queries. In evaluations conducted within the financial domain, TextRAG demonstrated a significant performance improvement, achieving an accuracy rate of 79% under a Shared Vector Store configuration, substantially outperforming the traditional benchmark accuracy of 19%. This highlights TextRAG’s exceptional capability in handling multi-document integration and complex information retrieval tasks.
-
EasyRAG:
EasyRAG is an innovative Retrieval-Augmented Generation (RAG) framework designed to simplify configuration, enhance integration, and optimize lightweight operations. It significantly lowers the learning curve for RAG applications through an intuitive interface and flexible module support, enabling developers to quickly and efficiently deploy and run complex natural language processing tasks while ensuring data security and high-performance processing. Whether for beginners or experienced developers, Easyrag offers a powerful and scalable solution for a wide range of application scenarios.