Gemma Scope 2

A new analysis warns that standard AI alignment techniques such as Reinforcement Learning from Human Feedback (RLHF) create a 'compliance mirage,' leaving large language models vulnerable to latent tr…

A replication of Anthropic's intentional control experiment on Gemma 3 27B Instruct found that the model has a stronger internal representation of a concept when told to think about it while writing a…