Nicholas Jiang

Interpretable Embeddings with Sparse Autoencoders: A Data Analysis Toolkit

Vision Transformers Don't Need Trained Registers

Interpreting and Editing Vision-Language Representations to Mitigate Hallucinations