为多变量时间序列模型提出幻觉检测与缓解新方法
Hallucination Detection and Mitigation with Diffusion in Multi-Variate Time-Series Foundation Models
- 用扩散模型评估多变量时间序列的幻觉水平
- 开源模型平均幻觉程度达弱基线的59.5%
- 所提方法可降低幻觉47.7%，提升模型可信度
自然语言处理领域的基础模型已有明确的幻觉定义及检测缓解方法，但多变量时间序列（MVTS）基础模型尚无相应标准。本文提出适用于MVTS的幻觉新定义，并基于扩散模型实现幻觉水平估计，构建了源自主流时序数据集的关系型基准数据集。实验表明，开源预训练MVTS填补模型的相对幻觉水平平均高达弱基线的59.5%；所提缓解方法可将其降低最多47.7%。该定义与方法有助于提升MVTS基础模型的可靠性与安全应用。
原文摘要 · Abstract (English)
Foundation models for natural language processing have many coherent definitions of hallucination and methods for its detection and mitigation. However, analogous definitions and methods do not exist for multi-variate time-series (MVTS) foundation models. We propose new definitions for MVTS hallucination, along with new detection and mitigation methods using a diffusion model to estimate hallucination levels. We derive relational datasets from popular time-series datasets to benchmark these relational hallucination levels. Using these definitions and models, we find that open-source pre-trained MVTS imputation foundation models relationally hallucinate on average up to 59.5% as much as a weak baseline. The proposed mitigation method reduces this by up to 47.7% for these models. The definition and methods may improve adoption and safe usage of MVTS foundation models.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。