Task-aware Retrieval with Instructions
Neural Search Talks β Zeta Alpha
00:00
Do You Think It's a Good Idea?
There's kind of like three main flavors or ways that people have been trying to do that so far, which is first the self-supervised approaches. Training on MS Marco and then transferring works pretty well, especially with the Cross- Strong Base Line. There's Prom applicator, which has some information about the task, kind of analogous to this. And then there's like the in-par's work where you just see the documents. So there's a little bit between methods in this.
Play episode from 03:47
Transcript
