Tag: #retrieval

한국어English日本語中文

← All posts

• The Structure of Practice: Why Practice That Goes Well and Practice That Lasts Are Different

  Spacing, retrieval practice, mixing things up. Three things with relatively solid evidence in learning research, set out together with their effect sizes. And then the central paradox of the field: the conditions that ma

  2026-08-16 · 14 min read #mindset#learning#practice#retrieval#spacing

• The Technique of Hallucinating Instead of Classifying, and How to Validate It — Why a Fake Label Beats the Raw Query

  Instead of putting a taxonomy of hundreds of entries into a prompt, have a small model invent a plausible fake classification and map it onto the real taxonomy with embeddings. Why that can work is explained by asymmetry

  2026-08-14 · 7 min read #llm#embedding#classification#search#retrieval

• The Claim of 100x Cheaper Is True Only When the Task Was Narrowed — Verification and Break-Even

  A case study published in August 2026 reports that a 4-billion-parameter-class open model, post-trained with reinforcement learning, matched frontier models on a retrieval task while cutting per-request cost by an order

  2026-08-09 · 8 min read #llm#cost#fine-tuning#retrieval#open-models

• When RAG Answers the Wrong Thing — a Debugging Procedure That Separates Retrieval Failure from Generation Failure

  When RAG returns a wrong answer, most teams start by fixing the prompt, but the majority of real causes sit in the retrieval stage. This post starts with the single experiment — hand the model the gold chunk directly — t

  2026-07-26 · 13 min read #llm#rag#retrieval#chunking#evaluation

• Production RAG Patterns — Why Naive RAG Fails and the Techniques That Actually Help

  A demo RAG system takes half a day to build, but the place it quietly breaks in production is almost always retrieval, not generation. This reference walks through chunking, embeddings and hybrid search (BM25 + vector),

  2026-07-11 · 9 min read #ai#rag#llm#retrieval#embeddings

• SOTA Text Embedding Models — The Heart of Search and RAG

  Text embeddings are the heart of search and RAG systems. We walk through contrastive learning and InfoNCE, dual encoders, hard negatives, the E5/BGE/GTE families, Matryoshka representation learning, and the MTEB benchmar

  2026-06-30 · 14 min read #ai-papers#embedding#retrieval#rag#contrastive-learning

• RAG Chunking Strategies: From Naive Splitting to RAPTOR

  Chunking determines 70% of your RAG system quality. This guide covers 5 chunking strategies with real code — from simple fixed-size splitting to Semantic Chunking, Parent-Child, and RAPTOR — with honest assessments of wh

  2026-03-18 · 8 min read #rag#chunking#text-splitting#langchain#raptor

• 1 Million Token Context Windows: Is RAG Becoming Obsolete?

  With Gemini 1.5 Pro's 1M token context and Claude's 200K, many developers ask: why bother with RAG? We compare real costs, latency, and quality to give you an honest answer.

  2026-03-18 · 6 min read #context-window#rag#llm#long-context#ai-development

• RAG Pipeline Production Guide: From Vector DB Selection to Chunking, Reranking, and Evaluation

  A comprehensive guide to building RAG (Retrieval-Augmented Generation) pipelines for production. Covers embedding model comparison, vector DB selection (Pinecone, Milvus, Weaviate, Qdrant, Chroma), chunking strategies, h

  2026-03-11 · 17 min read #llm#rag#vector-database#retrieval#embedding