用最优传输分析注意力分布，无监督检测机器翻译与摘要中的幻觉。
Layer-Resolved Optimal Transport for Hallucination Detection in NMT and Abstractive Summarization
- 基于跨注意力分布的几何距离检测幻觉，无需标注数据。
- 前四层注意力最敏感，第五层反向预测细微幻觉，正确翻译有探索性注意力模式。
- 适用于源文本脱节类幻觉，但对内容误述无效，适合关注可解释性的研究者。
最优传输（OT）已被证明可通过测量跨注意力分布与参考分布之间的几何距离，在无监督情况下检测神经机器翻译（NMT）中的幻觉。本文扩展分析至Fairseq DE-EN模型的六个解码器层（N=3,414），发现Wass-to-Unif与Wass-to-Data是互补的检测器，分别针对不同类型的幻觉；检测主要集中于第1至第4层，第5层对细微幻觉呈反向预测；而幻觉翻译缺乏正确翻译在首个解码步就存在的探索性注意力阶段。进一步评估该几何信号在摘要忠实度检测中的适用性：在AggreFact数据集上（N=1,116），无监督OT检测器在CNN/XSum上的平衡准确率分别为57.2%/57.6%，高于随机水平但显著低于监督方法MiniCheck-Flan-T5-L（69.9%/74.3%）。这一差距具有根本原因：不同于NMT中幻觉常表现为源文本脱离，不忠实摘要可能正确关注源词却扭曲其内容，此失败模式因构造上不可见于浓度型OT指标。在T5-base上的结构实验确认了解码器深度上一致的组织结构，第3层注意力集中度最高，第12层对生成质量最关键。综上，当故障模式为源文本脱离时，跨注意力上的OT是可靠检测工具，无论任务如何，均为原理清晰的可解释性分析手段，但在注意力下游发生的内容失真问题面前存在根本局限。
原文摘要 · Abstract (English)
Optimal transport (OT) has been shown to detect hallucinations in neural machine translation (NMT) by measuring the geometric distance between cross-attention distributions and a reference distribution, without any supervision. We extend this analysis to all six decoder layers of the Fairseq DE-EN model ($N=3{,}414$), showing that Wass-to-Unif and Wass-to-Data are complementary detectors specialised across hallucination types, that detection is concentrated in layers L1--L4 with L5 anti-predictive for subtler types, and that hallucinated translations lack the exploratory attention phase present in correct translations from the first decoding step. We further evaluate whether the geometric signal transfers to abstractive summarization faithfulness detection: our unsupervised OT detector on AggreFact ($N=1{,}116$) achieves $57.2\%$/$57.6\%$ balanced accuracy on CNN/XSum -- above chance but substantially below supervised MiniCheck-Flan-T5-L($69.9\%$/$74.3\%$). This gap is principled: unlike NMT hallucinations, unfaithful summaries can attend correctly to source tokens while misrepresenting their content, a failure mode invisible to concentration-based OT metrics by construction. Structural experiments on T5-base confirm consistent decoder organisation across depth, with Layer~3 showing peak concentration and Layer~12 being most critical for generation quality. Together, the results establish OT on cross-attention as a reliable detector when the failure mode is source disengagement, a principled interpretability tool regardless of task, and fundamentally limited when faithfulness failures occur downstream of attention.
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。