Tag
10 articles
Learn about SPARSEUP, a 149M-parameter sparse embedding model that sets new benchmarks for efficient retrieval systems. Understand how it leverages logit shift, top-k expansion, and case folding to achieve high performance while maintaining sparsity.
Google Research's Retrieve-for-Train (R4T) framework accelerates search results using reinforcement learning and diffusion models, achieving 12× to 20× faster query fan-out than traditional methods.
This article explains the advanced AI concepts behind video search technology, including multimodal learning, embedding generation, and scalable search algorithms that enable systems to index and retrieve terabytes of video content efficiently.
NEEDLE is a dynamic search benchmark that rebuilds its query set every hour to prevent answer leakage and provide realistic performance evaluation of search systems.
This article explains the technical challenges of AI knowledge systems through the case study of Elon Musk's Grokipedia, which has not been updated in months. It explores the architecture, maintenance requirements, and fundamental difficulties in creating reliable AI-generated encyclopedias.
This explainer explores how Meta's AI Mode works, combining retrieval-augmented generation with semantic search to extract answers from Facebook's vast user-generated content.
Learn to build a high-precision retrieve-and-rerank pipeline using the zeroentropy/zerank-2-reranker model, combining fast retrieval with advanced reranking for improved search quality.
Learn how to implement and compare BM25 and Retrieval-Augmented Generation (RAG) for information retrieval, understanding their fundamental differences in document ranking and response generation.
This explainer explores how AI-powered services like DeleteMe automatically detect and remove leaked personal data from the internet, examining the underlying technologies and challenges involved.
As language models gain the ability to process massive context windows, experts argue that selective retrieval methods like RAG remain more efficient and reliable than simply dumping all data into prompts.