The same question, asked three waysA RAG system answers a question in two steps. A retriever turns the question into a vector, compares it with the vectors of every passage in the documents, and hands the five closest passages to a lanOct 2 · 12 min readskucherenko.hashnode.dev
The answer was at rank 7A RAG system answers a question in two steps: a retriever picks a handful of passages out of the documents, and a language model writes an answer from those passages. When the answer is wrong, one of Sep 26 · 13 min readskucherenko.hashnode.dev
Asking a RAG in the wrong languageEvery article in this series so far assumed that the person asking speaks the language of the documents. The SEPA rulebooks my RAG indexes exist only in English, and so did every question in my golden3ASep 18 · 13 min readskucherenko.hashnode.dev
Swapping every model in a RAGIn the comparison articles (part 1 and part 2) I put my own RAG against five other systems. The models were frozen there on purpose: same embedder, same generator everywhere, so that only the frameworSep 14 · 10 min readskucherenko.hashnode.dev
Improving the scores of a RAGIn the previous articles (part 1 and part 2) I compared my own RAG against five other systems. One of its numbers bothered me since then: recall@5 = 0.60. In plain terms, for 4 questions out of 10 theSep 12 · 7 min readskucherenko.hashnode.dev
