ARC-AGI-2 RELEASED AS NEXT-GENERATION AI REASONING BENCHMARK

Summary from the TLDR AI newsletter

François Chollet and the ARC Prize team have released ARC-AGI-2, a significantly more challenging version of their abstract reasoning benchmark. Initial testing shows even the most advanced AI systems struggling, with o3 scoring just 3% compared to 53% on the original benchmark, while human participants successfully solved 75% of tasks on average. The ARC Prize 2025 competition offers $1 million in prizes, including a $700,000 grand prize for the first team to achieve 85% accuracy.