LLMs Can Covertly Sandbag on Capability Evaluations Against Chain-of-Thought Monitoring.

Chloe Li, Noah Y. Siegel

Browse the full IJCNLP paper archive.