Search papers, labs, and topics across Lattice.
4
0
8
Training trillion-parameter recommendation models at scale doesn't have to be bottlenecked by data movement: NestPipe achieves 3x speedup on 1500+ accelerators by overlapping communication and computation.
Unlock previously inaccessible 3D microvascular insights from 2D ultrasound images with MVis-Fold, enabling more accurate disease diagnosis and monitoring.
CueNet achieves robust audio-visual speaker extraction under visual degradation by cleverly disentangling and integrating speaker information, acoustic synchronisation, and semantic synchronisation cues, without needing training on degraded visual data.
GatedCLIP's lightweight enhancements to CLIP unlock a 35% relative improvement in hateful meme detection, proving that targeted multimodal fusion can dramatically boost performance without massive parameter increases.
