From 5e24dcb01cae0a0bef05646aff071c58c327fb36 Mon Sep 17 00:00:00 2001
From: Prashant Dixit <54981696+PrashantDixit0@users.noreply.github.com>
Date: Wed, 3 Jan 2024 17:50:56 +0530
Subject: [PATCH] blog link fix (#107)

---
 examples/Advance-RAG-with-HyDE/README.md | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/examples/Advance-RAG-with-HyDE/README.md b/examples/Advance-RAG-with-HyDE/README.md
index 9a142354..0f73dbe0 100644
--- a/examples/Advance-RAG-with-HyDE/README.md
+++ b/examples/Advance-RAG-with-HyDE/README.md
@@ -7,8 +7,11 @@ we provided Colab walkthrough for HyDE implementation   <a href="https://colab.r
 
 
 ### Learn deeper in Our Blog 
+The HyDE approach recognizes the difficulty of zero-shot learning and encoding relevance without labeled data. Instead, it leverages the power of language models and hypothetical documents. Here’s how it works:
 
-For a deeper dive into the cutting-edge technologies of HyDE, and to access detailed technical knowledge, check out our Medium Blog.
+1. **Generating Hypothetical Documents**: When a user enters a query, HyDE instructs a language model, like GPT-3, to generate a hypothetical document. This document is designed to capture relevance patterns but may contain inaccuracies.
+2. **Unsupervised Encoding**: The generated hypothetical document is then encoded into an embedding vector using an unsupervised contrastive encoder. This vector identifies a neighborhood in the corpus embedding space, where similar real documents are retrieved based on vector similarity.
+3. **Retrieval Process**: HyDE searches for real documents in the corpus that are most similar to the encoded hypothetical document. The retrieved documents are then presented as search results.
 
-[Read the Blog Post]()
+[Read the Blog Post](https://blog.lancedb.com/advanced-rag-precise-zero-shot-dense-retrieval-with-hyde-0946c54dfdcb)