🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Wrote a rather lengthy blog post on Muon It's still under construction and the notation may need more work (this is not my area of specialty exactly), but the main ideas should already be there I'm open to any feedback!

I view all companies in this space with a rather large pile of salt, but it's pretty hard for me to imagine someone other than cohere Cohere Labs doing this work and releasing this paper. I have a huge amount of respect for Sara Hooker Aidan Gomez et al.

If you're at #NAACL25, don't miss arlo_son presenting our paper on localizing MMLU to Korea! Session C: Oral/Poster 2: 2pm-3.30pm

Evaluating language models is hard and we are quite bad at it. That's why we are joining The University of Edinburgh and Hugging Face to organize a community coalition to focus efforts on improving the state of affairs.

Noam Brown “Designing hard evals is easy, designing meaningful ones is hard” -unknown

My least favorite part of writing dataset papers is figuring out what to cut out of the dataset description Alt. "my favorite part of writing dataset papers is writing the appendix"

Imagine this: An immigrant is trying to use an AI model to figure out her options regarding a complex legal status question. The AI company will "whistleblow" the immigrant's IP address and other PII to ICE. Masked men knocking on the door within the hour. The possibility of this

ACADEMICS: it is time to get our heads out of our *sses. This is not the moment for personal ambition, why your latest sophisticated widget beats rivals intricate theorem. The scientific franchise is under attack. It is time to defend it to the public. x.com/davidbau/statu…

Refuting your own paper after trying experiments at larger scales is incredibly based. Instant follow