Ai2 releases Tülu 3, a fully open-source model that bests DeepSeek v3, GPT-4o with novel post-training approach
DeepSeek-R1 released model code and pre-trained weights but not training data. Ai2 is taking a different approach to be more open.
DeepSeek Has More To Offer Beyond Efficiency: Explainable AI
DeepSeek-R1 charts a new path for AI through explaining its own reasoning process. Why does this matter and how will it benefit the world?
Did DeepSeek Copy Off Of OpenAI? And What Is Distillation?
The Medium post goes over various flavors of distillation, including response-based distillation, feature-based distillation and relation-based distillation. It also covers two fundamentally different ...
Rest of World3h
Non-western founders say DeepSeek is a proof that innovation need not cost billions of dollars
Entrepreneurs in Asia and Africa believe DeepSeek is proof that frugality and innovation can go hand in hand. DeepSeek’s open-source model has lowered the barriers for AI innovators outside the West.
Allen Institute for AI challenges DeepSeek on key benchmarks with big new open-source AI model
Amid the industry fervor over DeepSeek, the Seattle-based Allen Institute for AI (Ai2) released a significantly larger ...