Trust Functions: Near-Lossless Weak-to-Strong Generalization by Learning When to Trust the Weak Teacher

Part of

Machine Learning Model Evaluation and Robustness