Search papers, labs, and topics across Lattice.
2
0
3
0
Motion-as-Prompt reveals that enhancing motion reasoning in MLLMs can lead to substantial accuracy gains without altering model architecture or requiring retraining.
HUGIN boosts vision-language model accuracy for logistics sorting by over 15 percentage points, showcasing a significant leap in performance through innovative training techniques.
