RAG: Retrieval-Augmented Generation Explained скачать в хорошем качестве

RAG: Retrieval-Augmented Generation Explained

How do you ground a language model in real-world knowledge without retraining? RAG combines parametric and non-parametric memory into a single end-to-end architecture. ⏱️ TIMESTAMPS 0:00 The Knowledge Problem 0:50 Parametric vs Non-Parametric Memory 1:58 The RAG Architecture 3:09 RAG-Sequence vs RAG-Token 4:14 Training and Decoding 5:25 Open-Domain QA Results 6:35 Generation and Fact Verification 7:43 How Parametric and Non-Parametric Memory Cooperate 8:41 Ablations and Retrieval Analysis 9:44 Index Hot-Swapping: Updatable Knowledge 10:16 Impact and Legacy 📊 KEY RESULTS • State-of-the-art on 3 QA benchmarks with 18x fewer parameters than T5-11B • 42.7% more factual than BART on Jeopardy generation • Hot-swappable knowledge index - update without retraining 📄 RESOURCES Paper: https://arxiv.org/abs/2005.11401 Code: https://github.com/huggingface/transf... 👤 AUTHORS: Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, Douwe Kiela 🎬 ABOUT THIS VIDEO Created with LumoLab — transform research papers into explainer videos. Try it yourself: https://lumolab.org 📬 Subscribe for more AI research explainers! #AISafety #MachineLearning #ResearchExplained