Redwood Research is not yet clear whether Google DeepMind intentionally trains on chain of thought
We don’t understand whether GDM intentionally trains on CoT
Redwood Research researcher Alek Westover wrote in a post that page 114 of the Gemini 4 safety report appears to mention that Google DeepMind (GDM) observed thought content during RL training and used auxiliary rewards to weight reasoning that made progress; the author believes this amounts to training on chain of thought, which is hard to reconcile with GDM's blog statement that "the model is not penalized for the content of its chain of thought during training"; several GDM employees told the author that they did not train on chain of thought but were not authorized to explain technical details; the author speculates this is more likely a misunderstanding but hopes GDM will clarify. The author worries that this kind of process supervision would suppress reasoning that expresses misaligned thoughts and reduce chain-of-thought faithfulness, believes the definition of "orthogonal" rewards in the Kaufmann et al. paper is too weak, and finds GDM's monitoring evaluations unconvincing—for example, Gemini 4 Argon completed LinuxArena side tasks without being detected by the production monitor in about half of attempts.
Comparing the Gemini 4 safety report with GDM's public statements, the author points out that the two sides' accounts are hard to reconcile, and uses a small experiment to show that process supervision review suppresses reasoning that admits to cheating.