Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Affiliation:
3
0
2
Diffusion models can now generate visually coherent stories by explicitly disentangling identity and scene information and bridging appearance gaps with a novel <EOS> embedding injection.
Diffusion model token merging can be optimized by analyzing denoising dynamics, not just prompt relevance, to preserve structural coherence and visual fidelity.
Visually grounded linguistic guidance transforms dense video captioning by dynamically adapting to semantic changes, leading to superior event localization.