Search papers, labs, and topics across Lattice.
Email:
5
0
7
21
Instruction following learned in one modality transfers zero-shot to cross-modal editing tasks, enabling rich AV movie scene edits without paired training data.
LALMs can be easily tricked into "hearing" things that aren't there, with success rates as high as 95% on targeted attacks.
AudioChat tackles the complexity of "audio stories" by using LLM-driven tool-calling agents to simulate user interactions, enabling audio foundation models to generate, edit, and understand complex multi-source acoustic scenes.
Seamlessly extend and morph audio clips using a diffusion model with masked latents and classifier-free guidance, achieving near-realism and opening new creative possibilities for sound design.
A new model, TAC, uses synthetic training data to achieve state-of-the-art audio and audio-visual reasoning by generating temporally grounded captions that can then be fed into LLMs.
