|
|
|
|
|
|
by HarHarVeryFunny
285 days ago
|
|
|
I highly doubt they are doing that. If the quality is good then presumably it's also not the direct output of an LLM with RAG tool use. I'd guess it might be more of a structured/agentic approach - maybe having learnt how to map "search" strings to relevant data retrieval queries, then combining/summarizing the returned results. |
|
|