测量词汇训练数据覆盖对大型语言模型中幻觉检测的影响
Measuring the Impact of Lexical Training Data Coverage on Hallucination Detection in Large Language Models
AI总结：
本研究探讨词汇训练数据覆盖对大型语言模型幻觉检测的影响，通过构建后缀数组和n-gram统计，发现结合出现特征与对数概率可提升检测效果。
摘要：
大型语言模型（LLMs）中的幻觉现象是一个根本性挑战，尤其是在开放领域问答中。先前的研究尝试通过模型内部信号如token级熵或生成一致性来检测幻觉，而训练数据暴露与幻觉之间的联系仍被忽视。现有研究表明，LLMs在长尾知识上表现不佳，即生成答案的准确性会下降，对于在预训练中稀有的真实实体。然而，检查数据覆盖本身是否能作为检测信号却被忽视。我们提出一个补充问题：问题和/或生成答案的词汇训练数据覆盖是否为幻觉检测提供额外信号？为探讨这一点，我们构建了可扩展的后缀数组，覆盖RedPajama的1.3万亿token预训练语料库，以检索提示和模型生成的n-gram统计信息。我们在三个问答基准上评估其对幻觉检测的有效性。我们的观察表明，基于出现的特征单独使用时是弱预测因子，但与对数概率结合时能带来适度的提升，特别是在内在模型不确定性较高的数据集上。这些发现表明，词汇覆盖特征为幻觉检测提供了补充信号。所有代码和后缀数组基础设施均提供在https://github.com/WWWonderer/ostd。
英文摘要：
Hallucination in large language models (LLMs) is a fundamental challenge, particularly in open-domain question answering. Prior work attempts to detect hallucination with model-internal signals such as token-level entropy or generation consistency, while the connection between pretraining data exposure and hallucination is underexplored. Existing studies show that LLMs underperform on long-tail knowledge, i.e., the accuracy of the generated answer drops for the ground-truth entities that are rare in pretraining. However, examining whether data coverage itself can serve as a detection signal is overlooked. We propose a complementary question: Does lexical training-data coverage of the question and/or generated answer provide additional signal for hallucination detection? To investigate this, we construct scalable suffix arrays over RedPajama's 1.3-trillion-token pretraining corpus to retrieve $n$-gram statistics for both prompts and model generations. We evaluate their effectiveness for hallucination detection across three QA benchmarks. Our observations show that while occurrence-based features are weak predictors when used alone, they yield modest gains when combined with log-probabilities, particularly on datasets with higher intrinsic model uncertainty. These findings suggest that lexical coverage features provide a complementary signal for hallucination detection. All code and suffix-array infrastructure are provided at https://github.com/WWWonderer/ostd.
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。
