Xin Gao

Heterogeneous LLMs can share internal factual states without fine-tuning by exchanging word indices mapped to receiver-local vector directions.

Aesthetic code generation is unlocked by using a VLM judge to translate visual design principles from a reference image into actionable rewards for code models.

Multi-teacher distillation fails not because of bad routing, but because high-variance instruction feedback quietly drowns out precision domains like mathematics during shared student updates.

A novel PEFT method termed position anchor tuning (PAT) is proposed, which performs comparably to state-of-the-art methods while incurring significantly lower computational overhead and fewer trainable parameters.

A groundbreaking dataset and benchmark that bridges the gap between anatomical and metabolic analysis in whole-body PET/CT imaging, enabling advanced multimodal reasoning.

Safety judgments for reasoning traces are far more complex than for final responses, revealing critical gaps in current guardrail models' capabilities.

ViaMOBO reveals that understanding variable interactions can drastically enhance optimization efficiency in high-dimensional spaces, outperforming traditional methods.

FedTopo transforms federated learning by enabling reliable knowledge sharing through class relation topology, outperforming traditional methods in heterogeneous settings.

AI agents that understand social norms can outperform human-human interactions, achieving a 43% improvement in coordination.

Synthetic 4D cardiac MRI sequences can boost segmentation performance by over 1.4%, showcasing a powerful augmentation strategy for training robust AI models in medical imaging.