Wang Yang
2 indexed papers
Publications per year
Top categories
Frequent co-authors
Research Timeline
MOSAIC is a multi-objective framework that efficiently allocates a fixed supervised fine-tuning budget by turning failure profiles into actionable data mixtures, significantly improving model alignment across multiple safety and capability metrics.
The paper identifies a failure mode called spatial lexical bias in MLLMs, where adding a spatial word to options biases the model's choice, and demonstrates that this failure originates primarily from the language processing side rather than poor visual attention.
Papers
Mechanistic Diagnostics of Spatial Lexical Bias in Multimodal Large Language Model Spatial Reasoning
Chuang Ma, Qianying Liu, Tomoyuki Obuchi, Fei Cheng +5 more
The paper identifies a failure mode called spatial lexical bias in MLLMs, where adding a spatial word to options biases the model's choice, and demonstrates that this failure originates primarily from…