SeungJu Cha

Ditch the finetuning: this training-free method uses attention scores to generate rare concepts in images with more precision and control than LLM-guided approaches.

Diffusion models can now generate rare concepts and execute complex edits with greater fidelity, thanks to a training-free prompt blending technique that leverages statistical properties of the diffusion process itself.

LLM-augmented training with similarity-aware masking lets weakly-supervised video captioning models generate more accurate event descriptions and temporal boundaries, even with sparse training data.