LLM-as-a-Judge is increasingly used to evaluate policy responses on open-ended tasks that lack ground-truth answers. Existing work often directly converts th…
机构:上交大
来源:arXiv 2609.37145 | AI4Papers 论文推荐平台