I'm referring to the long-horizon task benchmark which has been exponential since GPT-2 https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-

I'm referring to the long-horizon task benchmark which has been exponential since GPT-2 https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/ https://metr.org/blog/2025-03-19-measuring-ai-ability-to-com...