Building Mini Minds: Our Adventures in Teaching Small LLMs to Reason

Join us for a candid exploration of our attempts to train small language models to solve puzzles through reinforcement learning. We'll share both the technical hurdles and melted GPUs as we try to replicate reasoning capabilities of models like OpenAI's o1 and Deepseek R1 at a smaller scale - all in pursuit of giving these tiny models their very own 'aha!' moments.