幻觉解毒：用于大语言模型训练的敏感性丢弃（SenD）
Hallucination Detox: Sensitivity Dropout (SenD) for Large Language Model Training
AI总结：
针对大语言模型幻觉问题，提出敏感性丢弃（SenD）训练协议，通过丢弃高方差嵌入索引降低训练方差，并开发高效特征值分数（EES）指标，提升模型可靠性达17%，且不影响下游性能。
摘要：
随着大语言模型（LLMs）日益普及，人们对其可靠性的担忧也在增加，尤其是由幻觉（即事实不准确或无关的输出）引起的问题。我们的研究探讨了训练动态中的不确定性与幻觉出现之间的关系。使用Pythia系列模型和多种幻觉检测指标，我们分析了幻觉趋势，并识别出训练过程中的显著方差。为解决这一问题，我们提出了敏感性丢弃（SenD），一种新颖的训练协议，通过确定性丢弃具有显著变异性的嵌入索引来减少训练期间的幻觉方差。此外，我们开发了一种无监督的幻觉检测指标——高效特征值分数（Efficient EigenScore, EES），它以2倍速度近似传统EigenScore。该指标被整合到我们的训练协议中，使SenD既具有计算可扩展性，又能有效降低幻觉方差。SenD将Pythia和Meta的Llama模型的测试时可靠性提升了高达17%，并在维基百科、医学、法律和编程领域增强了事实准确性，且不影响下游任务性能。
英文摘要：
As large language models (LLMs) become increasingly prevalent, concerns about their reliability, particularly due to hallucinations - factually inaccurate or irrelevant outputs - have grown. Our research investigates the relationship between the uncertainty in training dynamics and the emergence of hallucinations. Using models from the Pythia suite and several hallucination detection metrics, we analyze hallucination trends and identify significant variance during training. To address this, we propose Sensitivity Dropout (SenD), a novel training protocol designed to reduce hallucination variance during training by deterministically dropping embedding indices with significant variability. In addition, we develop an unsupervised hallucination detection metric, Efficient EigenScore (EES), which approximates the traditional EigenScore in 2x speed. This metric is integrated into our training protocol, allowing SenD to be both computationally scalable and effective at reducing hallucination variance. SenD improves test-time reliability of Pythia and Meta's Llama models by up to 17% and enhances factual accuracy in Wikipedia, Medical, Legal, and Coding domains without affecting downstream task performance.
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。
