Brian Cheung

Words That Make Language Models Perceive

Network of Theseus (like the ship)

Large Pretraining Datasets Don't Guarantee Robustness after Fine-Tuning

SecureLLM: Using Inference-time Compositionality to Build Secure Language Models for Private, Sensitive, and Secret Data

Look but Don’t Touch: Gradient Informed Selection Training

How Predictable is AI Progress?

MapQA: A Map-Question-Answering Benchmark for Visual Language Model Reasoning

Training the Untrainable: Introducing Inductive Bias via Representational Alignment

Training the Untrainable: Introducing Inductive Bias via Representational Alignment

ImageNet-RIB Benchmark: Large Pre-Training Datasets Don't Guarantee Robustness after Fine-Tuning

Let’s disagree to agree: Evaluating collective disagreement among AI vision systems