Yifeng Lu

Type-Compliant Adaptation Cascades

Larger language models do in-context learning differently

Simple synthetic data reduces sycophancy in large language models