Roei Herzig

Visualizing Thought: Conceptual Diagrams Enable Robust Planning in LMMs

DAVE: A VLM Vision Encoder for Document Understanding and Web Agents

Learning to Grasp Anything By Playing with Random Toys

Robotic Steering: Mechanistic Finetuning for Vision-Language-Action Models

Activation Reward Models for Few-Shot Model Alignment

Pre-training Auto-regressive Robotic Models with 4D Representations

Navigating the Labyrinth: Evaluating and Enhancing LLMs’ Ability to Reason About Search Problems