AnswerDotAI / byaldi

Use late-interaction multi-modal models such as ColPali in just a few lines of code.
Apache License 2.0
634 stars 62 forks source link

VisRAG Integration #46

Open cin-klein opened 1 month ago

cin-klein commented 1 month ago

VisRAG show impressive performance compare with ColPali on several benchmark. Could you integrate it @bclavie https://github.com/openbmb/visrag

bclavie commented 1 month ago

Planning to look into it next week! I'm off until Nov 4th and am getting heavy FOMO with all these announcements, but it is very cool!

dimroc commented 4 weeks ago

For anyone looking for the benchmarks, they can be found on page 7 of research paper: https://arxiv.org/abs/2410.10594.

@cin-klein, if you have any other benchmarks, I would love to see them.

bclavie commented 2 weeks ago

@cin-klein As an update, I'll be looking at adding VisRAG as it is cool and relatively low effort, but is not the top priority over QoL late-interaction updates as it's the original main purpose of the library.