LW - [Paper] AI Sandbagging: Language Models can Strategically Underperform on Evaluations by Teun van der Weij | The Nonlinear Library | Podwise