Hal Daumé III

Causal Differentiating Concepts: Interpreting LM Behavior via Causal Representation Learning

Natural Language Inference Improves Compositionality in Vision-Language Models

TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policies

Beyond Blanket Masking: Examining Granularity for Privacy Protection in Images Captured by Blind and Low Vision Users

DrM: Mastering Visual Reinforcement Learning through Dormant Ratio Minimization

$PREMIER-TACO$ is a Few-Shot Policy Learner: Pretraining Multitask Representation via Temporal Action-Driven Contrastive Loss