Search papers, labs, and topics across Lattice.
5
0
5
13
GoodDiffusion's innovative use of learnable signatures can thwart unauthorized access while preserving high-quality outputs for legitimate users, a game-changer for copyright protection in generative models.
ActiveScope achieves a remarkable 96.34% accuracy in high-resolution image understanding by actively correcting perception flaws in MLLMs.
Subtle and rare mistakes in egocentric videos can be detected with unprecedented accuracy by fusing insights from both coarse and fine action analyses.
Achieving nearly 50% Recall@1 in video retrieval without any training marks a significant leap in efficiency and effectiveness for complex user queries.
Naive combinations of diffusion and contrastive learning suffer from gradient conflict, but injecting contrastive signals from reconstructed images can jointly optimize discriminative and perceptual abilities in CLIP visual encoders.
