Validating the Comparison, Not Just the Scorer: Why Overall Accuracy Can Mislead in Interactive LLM Evaluation
arXiv:2610.31678v1 Announce Type: new Abstract: Many LLM studies compare how models respond to different follow-up prompts. For example, one prompt may neutrally ask a model to reconsider, while another insists that an invented film really exists. The difference is used to measure the effect of conversational pressure. An automatic scorer then labels whether each response makes an unsupported factual claim. Researchers commonly validate the scorer with one overall accuracy or $F_1$ score. Howeve
阅读原始来源
来源记录
研究分析
研究主张
论文标识：2610.31678。 来源描述：arXiv:2610.31678v1 Announce Type: new Abstract: Many LLM studies compare how models respond to different follow-up prompts. For example, one prompt may neutrally ask a model to reconsider, while another insists that an invented film really exists. The difference is used to measure the effect of conversational pressure. An automatic scorer then labels whether each response makes an unsupported factual claim. Researchers commonly validate the scorer with one overall accuracy or $F_1$ score. Howeve
评测与比较记录
可供比较的近期同类记录：“DS@GT ARC at CheckThat! 2026: A Verdict-Anchored Pipeline for Generating Fact-Checking Articles” (arXiv, 2026年11月5日)；“Same Origin, Different Verdicts: Text Production Conditions Confound Machine-Generated Text Detection in Russian” (arXiv, 2026年11月5日)。主题相近仅用于导航，未核验其是否属于同一需求或独立证据。
结论适用范围
结论只能在所述模型版本、任务、数据与推理预算内解释。社区票数、下载量或榜单位置不能证明另一工作流中的可靠性。
有效性与复现缺口
需核对样本选择、基线配置、不确定性、数据污染风险、负面结果以及代码或权重可获取性；此处没有本站复现实验记录。
复核路线
阅读论文或模型卡及评测附录，针对明确基线复现一个有限结论，再测试代表性用户任务并记录结论失效的场景。