通过融合多信号时序建模，提升文本生成中的幻觉检测精度。
Temporal Multi-Signal Fusion for Token-Level Hallucination Detection
- 基于文本统计、NLI蕴含与语言模型意外度的33维特征流，结合时序建模检测幻觉。
- 在RAGTruth数据集上达到0.840 AUC，较独立评分基线提升11点。
- 仅依赖生成文本和外部信号，适用于闭源模型且泛化能力强。
当前词粒度幻觉检测器对每个词独立评分，但在生成模型自信出错时失效。本文将幻觉视为时序延续的片段，通过序列标注方式检测：每个词从包含文本统计、自然语言蕴含（NLI）蕴含关系及语言模型意外度的33维特征流中得分，不依赖模型内部信息。使用双向门控循环单元（BiGRU）处理这些特征，在RAGTruth数据集上（10次随机种子）达到0.840 AUC，相比独立逻辑回归基线提升11点（p=0.002，Wilcoxon符号秩检验）。受控分解显示性能提升主要来自时序结构而非模型容量——证据在同段内由高置信位置向邻近模糊位置传播。该0.845上限在循环、状态空间（Mamba）和注意力架构中均重现，表明瓶颈在于特征集而非模型结构。因仅读取生成文本与外部信号，该检测器可应用于闭源模型，并在训练中未见过的语言模型生成文本上保持稳定，AUC下降不足4%。
原文摘要 · Abstract (English)
Token-level hallucination detectors score each token independently from a single signal, and fail exactly when the generating model is confidently wrong. This paper instead treats hallucination as a temporally extended span and detects it by sequence labeling: each token is scored from a 33-dimensional feature stream that fuses text statistics, Natural Language Inference (NLI) entailment, and language model surprisal, with no access to model internals. A Bidirectional Gated Recurrent Unit (BiGRU) over these features reaches an AUC of 0.840 on RAGTruth (10 seeds), an 11-point gain over an independent logistic-regression baseline (p = 0.002, Wilcoxon signed-rank). A controlled decomposition attributes most of the gain to temporal order rather than model capacity: evidence propagates from confident positions to ambiguous neighbors within a span. The same 0.845 ceiling recurs across recurrent, state-space (Mamba), and attention architectures, locating the bottleneck in the feature set rather than the model. Because it reads only the generated text and external signals, the detector works on closed-source models, and it keeps working on text produced by language models it never saw during training, losing under 4% AUC.
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。