标签
The paper introduces SurveyReview, a reviewer-aligned multi-dimensional benchmark for evaluating survey papers, along with SurveyAlign, a fine-tuned model that substantially improves alignment with human reviewers over prompt-based GPT-5.2 judging.