In this video, we review a new research paper from Microsoft Research that shows a new technique for training open-source models competitive with ChatGPT 3.5. This new technique allows small models (13b) to do more than imitate large foundational models. It allows them to learn the reasoning behind how to arrive at solutions. Enjoy :) Follow me on Twitter 🧠 - Subscribe to my Substack 🗞️ - Become a Patron 🔥 - Join the Discord 💬 - LLM Leaderboard - AGIEval - BIG-Bench Hard - Orca Paper - Imitation False Promise - FLAN Data -
Hide player controls
Hide resume playing