How will we do SFT on models with opaque reasoning?
Published on February 21, 2026 12:00 AM GMT Current LLMs externalize lots of their reasoning in human interpretable language.
Research
Academic or research source. Check the methodology, sample size, and whether it's been replicated.