Search papers, labs, and topics across Lattice.
6
0
5
Compact, automatically selected anatomical regions can drastically reduce hallucinations in medical VLMs without requiring expert annotations.
Fine-tuned behavioral models can achieve superior population-level alignment, closing the gap with general-purpose models in individual predictions.
ARTEMIS transforms weakly supervised video polyp segmentation by evolving reliable masks over time, achieving unprecedented accuracy in challenging clinical settings.
Systematic variability in expert annotations can be effectively modeled, leading to personalized segmentation outputs that outperform traditional methods.
Hallucinated tokens in LVLMs betray themselves through diffuse attention patterns and a failure to semantically align with any specific image region, enabling highly accurate detection.
Achieve state-of-the-art video polyp segmentation by adaptively selecting informative reference frames and aggregating multi-scale historical features with causal attention.
