Wenhua Zhao

collaborators

2 papers

cs.LG2025

Mitigating Think-Answer Mismatch in LLM Reasoning Through Noise-Aware Advantage Reweighting

Si Shen, Peijun Shen, Wenhua Zhao +1

Group-Relative Policy Optimization (GRPO) is a key technique for training large reasoning models, yet it suffers from a critical vulnerability: the \emph{Think-Answer Mismatch}, wh…

math.AG2024

A local to global question for linear functionals

George F. Seelinger, Wenhua Zhao

Let $F$ be an algebraically closed field and let $n \geq 3$ . Consider $V = F^{n}$ with standard basis ${e_{1}, \dots, e_{n}}$ and its dual space $V^*= {\mathrm{Hom}}_{F-{\mathr…