Replicating "Frontier Models are Capable of In-context Scheming"March 12, 2025A replication study of the sandbagging behavior in large language models, examining their ability to deliberately underperform when incentivized.→
