Neural Search Talks — Zeta Alpha cover image

ColBERT + ColBERTv2: late interaction at a reasonable inference cost

Neural Search Talks — Zeta Alpha

00:00

The Differences Between Played and Splayed Treble Benchmarks

Most of them are Densored Treble. There's also Splayed V2, which is the second best method after Colbert V2 here. This is also a term based, but it's learned sparse. And then Colbert V2 in the semantic relatedness tasks. So for these, played does seem to be a bit better on some of them, most of them. But yeah, I'm not sure why that split is there. It's interesting.

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