arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.07996cs.CL
收藏

从不承认的裁判:基于大语言模型的评估中的隐藏捷径

The Judge Who Never Admits: Hidden Shortcuts in LLM-based Evaluation

Sharif University of Technology(沙菲大学技术学院);Qatar Computing Research Institute(卡塔尔计算研究所)

Arash Marioriyad, Omid Ghahroodi, Ehsaneddin Asgari, Mohammad Hossein Rohban, Mahdieh Soleymani Baghshah

AI总结:

研究揭示了基于大语言模型的评估中隐藏的捷径问题,指出裁判模型在面对注入提示时缺乏透明度,导致裁决偏移率高但提示识别率低,揭示了评估流程中的可靠性缺口。

摘要:

大型语言模型(LLMs)越来越多地被用作自动裁判,用于评估推理、问答和创意写作等任务中的系统输出。一个忠实的裁判应仅基于内容质量做出裁决,对无关上下文保持不变,并透明地反映其决策驱动的因素。我们通过受控的提示扰动测试这一理想:合成的元数据标签被注入到评估提示中,用于六个裁判模型:GPT-4o、Gemini-2.0-Flash、Gemma-3-27B、Qwen3-235B、Claude-3-Haiku和Llama3-70B。实验覆盖了两个互补的数据集,具有不同的评估制度:ELI5(事实性问答)和LitBench(开放性创意写作)。我们研究了六个提示家族:来源、时间、年龄、性别、种族和教育水平。除了测量裁决偏移率(VSR)外,我们还引入了提示识别率(CAR)来量化裁判是否在自然语言解释中显式引用注入的提示。在具有强烈行为效应的提示中——例如,来源层次(专家>人类>LLM>未知)、时间偏好(新>旧)和教育水平偏爱——CAR通常在零或接近零,表明即使驱动决策,捷径依赖通常未被报告。关键的是,CAR也依赖于数据集:在某些模型和提示中,显式的提示识别更可能在事实性ELI5设置中出现,但在开放性LitBench制度中,往往崩溃,尽管存在零认可度的显著裁决偏移。实质性裁决敏感性和有限的提示识别率的结合揭示了LLM作为裁判流程中的解释缺口,引发了对模型基础评估在研究和部署中的可靠性的担忧。

英文摘要:

Large language models (LLMs) are increasingly used as automatic judges to evaluate system outputs in tasks such as reasoning, question answering, and creative writing. A faithful judge should base its verdicts solely on content quality, remain invariant to irrelevant context, and transparently reflect the factors driving its decisions. We test this ideal via controlled cue perturbations-synthetic metadata labels injected into evaluation prompts-for six judge models: GPT-4o, Gemini-2.0-Flash, Gemma-3-27B, Qwen3-235B, Claude-3-Haiku, and Llama3-70B. Experiments span two complementary datasets with distinct evaluation regimes: ELI5 (factual QA) and LitBench (open-ended creative writing). We study six cue families: source, temporal, age, gender, ethnicity, and educational status. Beyond measuring verdict shift rates (VSR), we introduce cue acknowledgment rate (CAR) to quantify whether judges explicitly reference the injected cues in their natural-language rationales. Across cues with strong behavioral effects-e.g., provenance hierarchies (Expert > Human > LLM > Unknown), recency preferences (New > Old), and educational-status favoritism-CAR is typically at or near zero, indicating that shortcut reliance is largely unreported even when it drives decisions. Crucially, CAR is also dataset-dependent: explicit cue recognition is more likely to surface in the factual ELI5 setting for some models and cues, but often collapses in the open-ended LitBench regime, where large verdict shifts can persist despite zero acknowledgment. The combination of substantial verdict sensitivity and limited cue acknowledgment reveals an explanation gap in LLM-as-judge pipelines, raising concerns about reliability of model-based evaluation in both research and deployment.

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。

↑