102481 - Here, Grok, you can see
N. Lygeros
Here, Grok, you can see the dataset structures of ARC 1 and ARC 2. They say that 100% of tasks have been solved by at least 2 humans (many by more) in under 2 attempts. Normally a smart person can solve 100% of tasks. So before the application of graph minors theory, we have to test them. But you are right the aim is to find forbidden minors. ARC 3 will include 1,000+ levels across 150+ environments that require agents to explore, learn, plan, and adapt. So there is a huge difference and this is beautiful for our approach.