Redwood Research questions whether Google DeepMind trains on Gemini 4 chains of thought
Overview
Redwood Research says Gemini 4's safety report indicates Google DeepMind uses auxiliary rewards that depend on chain-of-thought content.
The authors say this is concerning because judges could penalize misaligned reasoning. They say it conflicts with DeepMind's statement that it does not penalize chain-of-thought text, and they ask DeepMind to clarify.
Written by AI from the articles below · updated Nov 10, 4:07 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- Redwood Research BlogBlogRedwood Research questions whether Google DeepMind trains on Gemini 4 chains of thought
AIRedwood Research says Gemini 4's safety report suggests Google DeepMind uses auxiliary rewards that depend on chain-of-thought content, which it calls concerning because judges could penalize misaligned reasoning. The authors say DeepMind's statement that it does not penalize chain-of-thought text conflicts with the report, and they ask for clarification.
Heat Trend
Not enough continuous observations to show a trend yet.