Search papers, labs, and topics across Lattice.
4
0
7
0
Masking just 20 Visual Retrieval Heads in VLMs can lead to an 80-point drop in grounding accuracy, revealing their critical role in visual information extraction.
Self-SiMS redefines zero-shot video moment retrieval by eliminating reliance on noisy query similarities, achieving unprecedented accuracy in span generation.
Forget generic chatbot empathy – this agent proactively elicits and addresses negative thoughts with clinically-grounded CBT interventions.
Scaling up avatar pre-training to 1M in-the-wild videos unlocks emergent generalization capabilities like relightability and garment support, even without direct supervision.
