On January 20th, Deepseek took the world by storm by releasing R1: a Chain-of-Thought (CoT) Reasoning model competitive (and sometimes beating) OpenAI's o1-pro model and being completely free and open source.
Instead of getting lost in the hype and noise, we will be doing a deep, technical live broadcast led by Greg Schoeninger, an ex-IBM AI researcher and founder of Oxen.ai, this Friday diving deep into how to fine-tune R1 with synthetic data! This will NOT be a basic intro but a technically advanced paper dive into R1's math, how to generate synthetic data, and how to fine-tune the model yourself.
Who Are We and Why Are We Doing This?
We at Oxen.ai, believe that artificial intelligence should benefit all. Since the most powerful AI always starts with the best data, we are dedicated to building the best tools for iterating, tracking, and collaborating on data in any format.
We also host the arXiv Dives. A community of over 1,200 AI engineers, researchers, and other fellow nerds where we dive into the most recent and important AI paper releases to show you how the models were made and how you can use them in the real world.
Hope you can join us!
What People Are Saying About Deepseek R1:
https://x.com/gregschoeninger/status/1883741878098747901
https://x.com/karpathy/status/1884678601704169965
https://x.com/bgurley/status/1884695810799263977
See you Friday:)