Yilong Yang

GFR-SAM reveals that training-free segmentation can outperform traditional methods by effectively utilizing cross-image cues and contrastive filtering.

Many text-to-image models are safer than expected, but a subset poses significant risks that traditional evaluation methods fail to capture.

MLLMs can be manipulated to produce harmful outputs from benign inputs, exposing a critical vulnerability in their safety mechanisms.

Achieve state-of-the-art zero-shot camouflaged object segmentation by intelligently combining visual features, SAM, and MLLMs to overcome the limitations of relying solely on MLLMs for object discovery.