Data-augmentation

This paper explores the use of Test-Time Training (TTT) to enhance the abstract reasoning capabilities of large language models (LLMs), specifically focusing on the Abstraction and Reasoning Corpus (ARC) benchmark.