Table of Content
How Do Large Language Models "Think"? Non-Technical Notes on Lilian Weng's Why We Think
In-depth exploration of how the big language model "thinks"? Through non-technical reading notes of Lilian Weng's "Why We Think", you can appreciate its mystery. From Thinking in Tokens at the language layer to Thinking in Continuous Space at the structural layer, including the Chain-of-Thought mechanism, a variety of methods to improve accuracy, and different implementation methods. Not only...
Invoice recognition based on QwenVL2.5 model
Explore the secrets of invoice recognition based on the QwenVL2.5 model! This model can accurately extract key invoice information, such as invoice code, number, etc. From the performance data, the model size is positively correlated with the response time. The 3B parameter model is the fastest, followed by the 72B model. The 7B model is faster than the OCR-specific model, and the 72B model is...
An easy-to-understand introduction to MCP concepts
Want to learn more about RAG? Here is an easy-to-understand introduction to the concept of RAG. Recently, open source big models have become very popular, and the Model Context Protocol (MCP) has attracted much attention. MCP is like a USB interface for AI applications. It standardizes the way context is provided. It is a bridge connecting AI applications with different data sources, and can...
RAG (Retrieval Enhanced Generation): The Ultimate Guide to Improve Performance of Large Language Models
Explore RAG (Retrieval Enhanced Generation) The ultimate guide to improving the performance of large language models! It can effectively solve the dilemma of la...
RAG Comprehensive Guide: Late Chunking vs Contextual Retrieval Solves Contextual Problems
Deeply exploring RAG technology, it is not as simple as you think. The recent hustle and bustle of Agents has caused its popularity to decline, but it is unknow...
Building Production-Grade RAG Pipelines Based on Gemini and Qdrant: Design Guidelines and Code Practice
Deeply explore the secrets of building production-grade RAG pipelines based on Gemini and Qdrant! This article elaborates on the core values and application s...
The Future of Knowledge Q&A: Breaking Through the Limitations of Traditional RAG
Deeply explore the ultimate form of knowledge Q&A and break through the limitations of traditional RAG. Although traditional RAGs use vector knowledge base and...
Three Stages of AI Agent Implementation in the Enterprise
In-depth discussion on the key stages of AI agents' implementation in the enterprise, focusing on RAG. Understanding its basic implementation method combining s...
REST API Migration to MCP Server: A Breakthrough for Traditional Enterprise IT Applications
A deep discussion on the migration of REST API to MCP Server is a breakthrough for traditional enterprise IT applications. The MCP launched in November 2024 is...
Optimizing Knowledge Graph and LLM Interfaces: Breaking Through the Performance Bottleneck of Complex Reasoning
In-depth discussion on the optimization of knowledge graphs and LLM interfaces, breaking through the bottleneck of complex inference performance! Understand the...