Yoav Goldberg: Frontier Model Reasoning Is Not "Emergent"

yoavgo · x · 2026-09-25

AI researcher Yoav Goldberg clarified in a discussion with Barak Rotblat that he isn't questioning reasoning in general, but the specific kinds and levels of reasoning shown by current frontier models.

He explicitly does not think these abilities are "emergent" — though definitions of the term may differ. The thread started from his admission that he has no mental model for how such reasoning traces arise from RL training, suspecting massive human-produced SFT data.

Related event: Yoav Goldberg Questions Whether Frontier LLM Reasoning Is Truly Emergent, Sparking Researcher Debate(8 posts)→

Original post →

More from AGI Musings

AGI Musings channel →