Synthetic document finetuning for instilling positive traits

This is the fifth in a series of informal research updates from the Google DeepMind Language Model Interpretability team, in interpretability and adjacent areas. The fourth post can be found here. TLDR: Via adapting the methods of Marks et al and Li et al, we train Gemini 3 Flash to have certain tra

Related Stories

How questions in the quiz are more difficult and why can you not try it again?

AI adoption accelerates in India, yielding benefits despite infrastructure challenges

European Commercial Space Venture Hits Another Launch Roadblock

A 13-word Reddit comment can trick AI search into recommending scams, researchers find