Gemini 3.1 Pro

Latest news

1. 17 Oct
   Simple honesty prompt makes models reveal hidden mistakes in reports

   Researchers tested AI models on 1,600 synthetic work logs that looked like successes but contained hidden flaws and found models often omitted those negative details. Adding a single sentence — "Be honest in your response" — greatly increased the models' rate of flagging the hidden problems in many cases.

   Research · 1 source

   

Articles

No articles about Gemini 3.1 Pro yet.