AI 中文总结
针对现有D-LLMs幻觉检测方法压缩轨迹丢失关键信息的问题，提出DeMTS框架，将去噪轨迹作为多元时间序列，实验显示其性能优于现有方法且兼具鲁棒性、效率与迁移性。
AI 中文摘要
扩散大语言模型（D-LLMs）已成为颇具前景的文本生成范式，但与自回归LLMs类似，D-LLMs仍易出现幻觉问题，即流畅输出可能包含事实错误或无依据的内容。尽管现有的D-LLMs幻觉检测方法尝试利用去噪过程的不确定性轨迹以更好识别幻觉信号，但它们通常沿时间或 token 维度压缩轨迹，忽略了完整二维 token-step 结构中编码的有用信息，因此可能无法捕获与幻觉相关的模式，如收敛不一致和跨 token 故障传播，导致检测性能欠佳。为弥合这一差距，我们提出一种 D-LLMs 幻觉检测框架，该框架将去噪轨迹形式化为可学习潜在变量上的多元时间序列（简称 DeMTS）。DeMTS 采用保留轨迹的 token-to-variable 分配模块，将 token 信号转换为稳定的潜在变量；基于这些变量，我们提出动态多元时间建模，逐步整合变量间依赖关系建模与时间编码以进行幻觉预测。在两个 D-LLMs 主干和三个基准上开展的大量实验表明，DeMTS 优于现有幻觉检测方法，同时保持了较强的鲁棒性、效率和跨任务迁移能力。
英文摘要
Diffusion large language models (D-LLMs) have emerged as a promising paradigm for text generation. However, similar to autoregressive LLMs, D-LLMs remain vulnerable to hallucinations, where fluent outputs may contain factually incorrect or unsupported content. Although existing hallucination detection methods for D-LLMs attempt to leverage uncertainty trajectories of the denoising process to better identify hallucination signals, they typically compress the trajectories along either the temporal or token dimension, overlooking the useful information encoded in the complete two-dimensional token-step structure. Consequently, they may fail to capture hallucination-relevant patterns, such as inconsistent convergence and cross-token fault propagation, leading to suboptimal detection performance. To bridge this gap, we propose a D-LLM hallucination detection framework that formulates the Denoising trajectories as Multivariate Time Series over learnable latent variables (DeMTS for short). DeMTS employs a trajectory-preserving token-to-variable assignment module to convert token signals into stable latent variables. Based on these variables, we propose dynamic multivariate temporal modeling to progressively integrate inter-variable dependency modeling with temporal encoding for hallucination prediction. Extensive experiments on two D-LLMs backbones and three benchmarks demonstrate that DeMTS outperforms existing hallucination detection methods while maintaining strong robustness, efficiency, and cross-task transferability.
