Articles
Loop Engineering for RAG: Small Loops Within Steps, Big Loops Across the Pipeline⭐9
Loop engineering in RAG: detecting failures, adjusting, and retrying steps to turn brittle one-shot pipelines into resilient systems that recover and deliver.
RAG Workflow and Loop Engineering: The Dispatcher That Decides When to Loop and When to Stop⭐9
A dispatcher pattern for RAG pipelines that decides when to loop and stop, coordinating retrieval, aggregation, and synthesis for enterprise reliability.
Before Full Agentic RAG: Know How You Decide, and the Parsing Methods You Pick From⭐9
Before full agentic RAG, control parsing choices with an explicit dispatcher: plan, execute, and log each method for transparent, validated enterprise retrieval...
Simulator-Grounded Large Language Models for Industrial Causal⭐8
Compare three methods for grounding LLMs in a wastewater simulator, achieving up to 99.5% causal QA accuracy and robust cross-plant transfer.
Prompt, Context, Loop: The Three Engineering Layers Every RAG⭐10
Explore the three engineering layers of RAG systems: prompt, context, and loop. Learn how each layer works and why their evolution isn't a simple sequence.
How I Reproduced BM25, Dense Retrieval, and SPLADE on a 16GB MacBook⭐9
I reproduced BM25, dense retrieval & SPLADE on a 16GB MacBook, matching RAG baselines; includes crash fixes & score interpretation.
RAG-Anything Tutorial: Build a Multimodal Retrieval Pipeline for⭐8
Build a multimodal retrieval pipeline in Google Colab with RAG-Anything. Learn to index text, tables, equations, and images for hybrid search and AI-powered rea...
MosaicLeaks: Can Your Research Agent Keep a Secret?⭐9
MosaicLeaks exposes a privacy flaw in AI research agents, allowing attackers to combine harmless queries to reconstruct sensitive data from HR, finance, and hea...
Dr-DCI: Scaling Direct Corpus Interaction via Dynamic Workspace⭐9
Dr-DCI scales direct corpus interaction by dynamically expanding workspaces via retriever-steered actions, achieving 71.2% accuracy in Browsecomp-Plus benchmark...
