Search papers, labs, and topics across Lattice.
Singapore University of Technology and Design
6
0
7
DART bridges the reasoning gap in zero-shot video temporal grounding, achieving a remarkable 3.5-point mIoU improvement with over 7 times fewer frames.
Visual-Seeker outperforms proprietary models by actively engaging with visual details, redefining multimodal search capabilities.
Ditching left-to-right thinking boosts video understanding: Masked diffusion models let AI revise its guesses about when actions happen in videos, leading to more precise timing.
Unsupervised visual tracking gets a surprising boost from text-to-image diffusion models, which can be prompted to highlight the target object in each frame without any training data.
Models trained on standard UAV footage can fail dramatically when deployed on drones with different camera angles, revealing hidden viewpoint biases.
Current multimodal browsing agents are surprisingly bad at using visual information on webpages, with even top models scoring below 50% accuracy on a new visual-native search benchmark.
