Towards eliciting latent knowledge from LLMs with mechanistic interpretability

Re-assign community