<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Kushal Lab</title><description>Every post is an experiment. Test the test before you test the model.</description><link>https://blog.seemakurthi.com/</link><item><title>HyDE: Embed the Answer You Wish You Had</title><link>https://blog.seemakurthi.com/exp/hyde-hypothetical-document-embeddings/</link><guid isPermaLink="true">https://blog.seemakurthi.com/exp/hyde-hypothetical-document-embeddings/</guid><description>When your query and its answer live in different parts of the embedding space — and how generating a hypothetical answer first bridges that gap.</description><pubDate>Wed, 27 May 2026 00:00:00 GMT</pubDate></item><item><title>Step-Back Prompting: Retrieve the Principle Before the Detail</title><link>https://blog.seemakurthi.com/exp/step-back-prompting-retrieval/</link><guid isPermaLink="true">https://blog.seemakurthi.com/exp/step-back-prompting-retrieval/</guid><description>Before answering a specific question, ask a broader one — retrieve foundational context, then answer the original with both in hand.</description><pubDate>Wed, 27 May 2026 00:00:00 GMT</pubDate></item><item><title>CRAG: Grade Your Retrieval Before You Trust It</title><link>https://blog.seemakurthi.com/exp/crag-corrective-rag/</link><guid isPermaLink="true">https://blog.seemakurthi.com/exp/crag-corrective-rag/</guid><description>Retrieval can fail silently. CRAG adds a confidence-scoring evaluator between the retriever and the generator — discarding bad chunks before the LLM ever sees them.</description><pubDate>Wed, 27 May 2026 00:00:00 GMT</pubDate></item><item><title>Vector Spaces &amp; Cosine Similarity: The Math Behind Dense Retrieval</title><link>https://blog.seemakurthi.com/exp/vector-spaces-cosine-similarity/</link><guid isPermaLink="true">https://blog.seemakurthi.com/exp/vector-spaces-cosine-similarity/</guid><description>Why two texts can mean the same thing without sharing a single word — and how a dot product turns that intuition into a retrieval engine.</description><pubDate>Sat, 23 May 2026 00:00:00 GMT</pubDate></item></channel></rss>