Neural Search Talks — Zeta Alpha cover image

ColBERT + ColBERTv2: late interaction at a reasonable inference cost

Neural Search Talks — Zeta Alpha

00:00

Interaction Base and Representation Based Winter Interaction

We see Colbert is competitive with Burnt-based, maybe slightly worse. These authors train it, it seems a little bit better than another row they cite. So not quite as good re-ranking, but already quite close and much more efficient. If you look at latency, this is 61 milliseconds versus like 10 seconds. Yeah. That's that's very interesting. All right. We should jump to the most interesting or most relevant results. I think maybe you can look at table two maybe.

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