Jan and Daniel share their journey to winning the ARC Challenge, detailing their innovative approach using a large language model with a unique tokenization strategy. They discuss the importance of test time training and the challenges faced during the competition, highlighting how iterative improvements led to significant performance gains.
