Frontier LLMs drop from 83% to 43% once reasoning has to chain across domains

1 comments

• westurner 12 hours ago ago

  Is that the model, all models, the agent?

  Does something like this make a difference?

  "Schema Harness Achieves ~99% on Arc‑AGI‑3 Public" (2026-07) https://news.ycombinator.com/item?id=48935905