AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
How to Engineer a Search Engine With Dense Retrieval Signals
The big challenge that people have with, I mean, just learning to rank is extroctwork. And do you also think maybe a connected question to think that we will marry, you know, the dense retrieval signals with learning to rank in some way? Does it make sense? Yeah, yeah. It makes sense because one of the biggest problems with sort of training data is it's just horribly biased towards whatever the old algorithm is. But if we just look at like the top n, the top 10 or so, what's actually getting clicks, we might be able to start to learn about what differentiates them.