arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2610.27247cs.CLcs.AI
收藏

面向大语言模型的幻觉压力测试的目标感知筛选

Target-Aware Curation of Hallucination Stress Tests for Large Language Models

Dartmouth College(达特茅斯学院);Oakland University(奥克兰大学)

Jiho Lee, Bingquan Zhang, Lin Ting Liang, Soroush Vosoughi, Weicheng Ma

AI总结:

本研究提出一种目标感知的大语言模型幻觉压力测试筛选方法,通过迭代编辑基于来源的种子提示,经实验验证该方法可显著提升幻觉严重程度评定,筛选后的提示能在不同模型间复用且具备诊断价值。

摘要:

幻觉压力测试可能在不同语言模型上产生相似的平均分数,即使不同提示会在每个模型中引发不同的幻觉。我们开发了一种程序,该程序迭代编辑基于来源的种子提示,以暴露目标模型中的幻觉。一个大语言模型编辑器会根据目标模型的回答以及评判者的特定类别分数和解释来修改每个提示。在主要设置中,目标模型还会收到关于其先前响应的反馈。实验涵盖了来自七个数据集和十个主题的500个种子提示,在开放权重模型和闭源模型上评估了内在和外在幻觉。与种子提示相比,编辑后的提示显著提高了评判者评定的严重程度。在基线配置中,针对外在幻觉的运行中,77.2%的保留响应属于评判者的强幻觉类别;针对内在幻觉的这一比例为74.2%。后续编辑轮次贡献了平均分数增长的一半以上。在通过模型辅助来源有效性检查的219个提示中,有89个会引发包含无根据主张的响应。在五个模型中,200个筛选后的提示产生了相似的平均严重程度分数,尽管提示级分数之间的成对斯皮尔曼相关系数仅为ρ=0.0387至0.1897。筛选后的提示可在不同模型间重复使用,但其诊断价值取决于来源有效性以及它们在每个模型中暴露的失败情况。

英文摘要:

Hallucination stress tests can yield similar average scores across language models even when different prompts elicit hallucinations in each model. We develop a procedure that iteratively edits source-grounded seed prompts to expose hallucinations in a target model. An LLM editor revises each prompt in response to the target's answer and a judge's category-specific score and explanation. In the main setting, the target also receives feedback on its previous response. Experiments cover 500 seeds from seven datasets and ten topics, with intrinsic and extrinsic hallucinations evaluated across open-weight and closed-source models. Editing substantially increases judge-rated severity over the seeds. In the baseline configuration, retained responses fall in the judge's strong-hallucination bin in 77.2% of runs targeting extrinsic hallucination and 74.2% targeting intrinsic hallucination. Later editing rounds account for more than half of the mean score increase. Among 219 prompts that pass model-assisted source-validity checks, 89 elicit responses containing unsupported claims. Across five models, 200 curated prompts yield similar average severity scores, although pairwise Spearman correlations between prompt-level scores are only $ρ=0.0387$ to $0.1897$. Curated prompts can be reused across models, but their diagnostic value depends on source validity and the failures they expose in each model.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑