HumbleBench: Measuring Epistemic Humility in Multimodal LLMs

Grounded Chain-of-Thought Makes Multimodal LLMs More Data-Efficient

Fine-Tuning 13B LLM or Stable Diffusion 3.5 Large Within a Single 24GB GPU

Watermarking Autoregressive Image Generation Models

Mitigating Shortcuts in Visual Reasoning with Reinforcement Learning