Yuqing Zhang

Current MLLMs struggle with creative decoding, achieving only 50.7% accuracy in understanding cross-concept relations, revealing a critical gap in their cognitive capabilities.

SeamGen learns from professional seam layouts to generate UV seams that artists actually prefer, outperforming traditional methods in both quality and alignment.

Human-like color categories can emerge in neural agents by simply upsampling rare color terms and training with multiple listeners, closing the gap between artificial and human color lexicons.

LLMs, when combined with rule-based analysis and iterative refinement, can significantly outperform existing tools in detecting subtle smart contract vulnerabilities arising from library misuse.

LLMs can boost code clone detection accuracy by selectively arbitrating only 0.2% of uncertain cases flagged by a multimodal fusion model, achieving a 0.3% absolute Macro-F1 gain.