ARC-AGI

The ARC-AGI benchmark is a significant measure in the field of artificial intelligence, focusing on an AI's general reasoning capabilities. Recently, there has been a notable achievement where GPT-4o reached a 50% score on the ARC-AGI benchmark, surpassing the previous best score of 34%. This benchmark involves several examples and problems that require the system to infer rules and output correct results corresponding to the problem diagram.