Zhuo Wang

Learning a residual correction to a physics-informed estimate dramatically boosts QNN performance in cloud-cover parameterization, outperforming both direct QNNs and compact MLPs.

Continuous interaction with LLMs can skew grading standards, leading to systemic unfairness in academic assessments.

Forget expensive CoT curation: CoTEvol's genetic evolution approach boosts correct-CoT synthesis by 30% and significantly improves LLM math reasoning performance.