Armen Aghajanyan

When Worse is Better: Navigating the Compression Generation Trade-off In Visual Tokenization

Small Molecule Optimization with Large Language Models

MoMa: Efficient Early-Fusion Pre-training with Mixture of Modality-Aware Experts

DOMINO: A Dual-System for Multi-step Visual Language Reasoning

Jointly Training Large Autoregressive Multimodal Models

Eliciting Attributions from LLMs with Minimal Supervision