Comment by spacedoutman

This research is useless and nearly all other LLM research is too.

gpt 5.2 is the strongest model they tested, a nearly 6 month old model.

Traditional research can not keep up.