3 accepted papers
As language models grow more capable, accurate capability evaluation becomes essential for safety decisions. If models can deliberately underperform on dangerous capability evaluations---a behavior known as \emph{sandbagging}---they may evade safety measures designed for their true capability level.…
Current Large Language Models (LLMs) benchmarks are often based on open-ended or close-ended QA evaluations, avoiding the requirement of human labor. Close-ended measurements evaluate the factuality of responses but lack expressiveness. Open-ended capture the model’s capacity to produce discourse re…
As AI models are deployed with increasing autonomy, it is important to ensure they do not take harmful actions unnoticed. As a potential mitigation, we investigate Chain-of-Thought (CoT) monitoring, wherein a weaker trusted monitor model continuously oversees the intermediate reasoning steps of a mo…