Yoav Goldberg: Frontier Model Reasoning Looks Like a Training Process Question, Not a Black Box
yoavgo · x · 2026-09-25
NLP researcher Yoav Goldberg discussed how to think about the specific kinds and levels of reasoning shown by current frontier models. He argues the question is closer to "what is the training routine of chess masters" than "how does the human brain play chess": the answer likely lies in aspects of the model creation process not yet fully understood, rather than in some mysterious latent process. He also rejects calling such abilities "emergent," noting the disagreement may be definitional.
More from Research
- AVIS paper accepted at NeurIPS 2026: jointly scaling visual context and reasoning per query — CSProfKGD · 2026-09-25
- ICLR hits 62,000 abstract submissions as NeurIPS decisions drop a day before deadline — yoavgo · 2026-09-25
- Weco preprint: agent self-improves its own software 7 times over 8 days, gains transfer to 4 benchmarks — VraserX · 2026-09-25
- Neurosymbolic researchers argue LLMs can't do abduction alone — luislamb · 2026-09-25
- Open-Jev multilingual text classification dataset trends on Hugging Face — ZefanCai · 2026-09-25
- New Work Formalizes Continual Learning's Stability-Plasticity Trade-off via Probability Paths — hbouammar · 2026-09-25