Seong-Whan Lee

Structured multi-branch reasoning in T2I-ICL can dramatically enhance image generation consistency and semantic alignment, outperforming existing prompting strategies.

Direct composition learning in DiCE-CIR achieves state-of-the-art zero-shot image retrieval without the overhead of projection methods, streamlining the process significantly.

Overparameterized PINNs can self-partition into ineffective modules, but a new training framework restores robust learning and achieves unprecedented accuracy.

Language-aware episodic training can significantly enhance speaker verification performance by disentangling speaker identity from language-specific features.

VibE-SVC2 achieves unprecedented independent control over singing styles, including pitch and timbre, while effectively addressing the challenges of pitch-energy entanglement and subharmonic characteristics.

By explicitly modeling relationships between multiple relevant video segments, ClipTBP significantly improves video moment retrieval, especially when queries are ambiguous.

Injecting nonverbal cues like laughter and sighs into speech synthesis is now more expressive and natural, thanks to a novel training strategy that overcomes data scarcity.