News for hackers. Stuff that matters.

A comment on Gemini 3 Deep Think. Parent comment

Re:Gemini 3 Deep Think by matjet on Friday February 13, 2026 @06:22AM (#47001509)
Look what they need to mimic a fraction of [the power of having the logit probabilities exposed so you can actually see where the model is uncertain]
[ Parent | on HN ]
Re:Gemini 3 Deep Think by kfajdsl on Friday February 13, 2026 @05:37PM (#47008792)
All the LLM logprob outputs I've seen aren't very well calibrated, at least for transcription tasks - I'm guessing it's similar for OCR type tasks.
[ Parent | on HN ]
Re:Gemini 3 Deep Think by energy123 on Saturday February 14, 2026 @05:11AM (#47013297)
"I already decided in my private reasoning trace to resolve this ambiguity by emitting the string '27' instead of '22' right here, thus '27' has 100% probability"
[ Parent | on HN ]