Neural Search Talks — Zeta Alpha cover image

ColBERT + ColBERTv2: late interaction at a reasonable inference cost

Neural Search Talks — Zeta Alpha

00:00

The Future of Dense Retrieval

There's a lot of research and so many people who are focusing on that mode of dense retrieval, of like single vector to single vector. I think it has a few attractive properties. One is that you can just use an approximate nearest neighbor index with it directly. You don't have to do this Colbert,. Like re-ranking, re-ranking with Central Aid, whatever the stuff. That's kind of nice. People like the simplicity or what they feel is the simplicity of just having one vector for the document.

Transcript
Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.
App store bannerPlay store banner