Search papers, labs, and topics across Lattice.
6
0
5
A training-free self-guidance method can dramatically enhance output diversity in flow models without the overhead of external reward systems.
Achieving a 4x speedup in multi-reference image generation without sacrificing visual quality by intelligently dropping reference tokens.
Transforming 3D image editing from ambiguous controls to precise geometry specifications, this method allows for unprecedented accuracy in real-world applications.
Exploiting temporal redundancy in video tokenization can yield a staggering 31x speedup in inference time while maintaining high reconstruction fidelity.
Finally, text-to-image generation can handle complex 3D scenes with realistic occlusions, thanks to a clever translucent box representation that explicitly encodes hidden object regions.
Text-to-image models consistently depict countries like India and Nigeria as disproportionately impoverished, revealing stark socio-economic biases that GeoDiv can now systematically measure.
