After a change to a RAG system, the answers still read fluently. Whether they got better or worse is another matter: "The contract cannot be terminated with 30 days' notice" shares almost every word with the correct answer and says the opposite.Third post on our Baltic DB&IS 2026 paper on Graip.AI Agent Labs (with Guntis Bārzdiņš): software tests check that it works, LLM judges check the answers, offline and online.https://emukans.github.io/posts/rag-evaluation/#RAG #LLM #Evaluation
After a change to a RAG system, the answers still read fluently. Whether they got better or worse is another matter: "Th...
