And we tested that, comparing the answers of VLMs with those of more than 1500 volunteers and 1.5M classifications of almost 46000 galaxy candidate images in a Zooniverse campaign.
The results? Well, VLMs are good, but cannot fully reproduce the "wisdom of crowd"!
https://t.co/kQp3nCwOfJ
New week, with a paper...!!
We asked a simple question: Do Vision-Language models "see" dwarf galaxy images the way we do? (1/2)
Can frontier language model agents replicate astrophysics research papers? Clearly not yet -- but models are slowly getting better! Excited to finally put out ReplicationBench, the work of an awesome team of astrophysicists from across Stanford's KIPAC, SLAC, and C4DU.
Super excited to be speaking alongside giants such as @ylecun and @Yoshua_Bengio at the world model workshop 🚀 at Mila!
Hope to see many of you in the wondrous Montreal!
Are you a PhD student or postdoc looking to plug into a network of innovative data science opportunities?
Join us at the Rising Stars in Data Science Workshop, hosted by @Stanford, @UCSanDiego, and @UChicago!
Learn more at an info session on Thursday, July 17th
⬇️Link⬇️
What if we found a way to tell the personality of an AI system? ✨
Together with @Pranav_AL, we're happy to introduce Supernova: https://t.co/FyEqHuRgHQ🍓.
Because stories matter. And this is a different kind of story - check out the incredible thread below made by the one and only and my wonderfully inspiring fellow explorer @Pranav_AL 🚀.
P.S. We'll be at ICML, come talk to us!