SpanUQ：面向大语言模型生成的 Span 级不确定性量化
SpanUQ: Span-Level Uncertainty Quantification for Large Language Model Generation
研究者提出 SpanUQ，一种轻量级探针，将多采样推理中的不确定性知识蒸馏到对 LLM 隐状态的单次前向传播中，实现 Span 级不确定性估计。该方法采用 DETR 风格 span 解码器与 Beta 混合分布，在五个 LLM 主干上取得最佳 span 级不确定性质量，比最强探针基线和所有采样方法快 10~20 倍，span 检测器 F1 达 0.910。
Authors:Yimeng Zhang, Yingying Zhuang, Ziyi Wang, Yuxuan Lu, Pei Chen, Aman Gupta, Zhe Su, Ming Tan, Zhilin Zhang, Qun Liu, Manikandarajan Ramanathan, Rajashekar Maragoud, Edward Vul, Jing Huang, Dakuo Wang
Abstract:Uncertainty estimation is essential for the trustworthy deployment of LLMs: it underpins self-refinement and serves as a decision signal for downstream systems, from verifiers to collaborating agents. However, existing approaches operate at suboptimal granularities: token-level scores lack semantic coherence, while sequence-level scores fail to localize errors. We formalize Span-Level Uncertainty Estimation (SLUE), a new task that targets the natural granularity for uncertainty: semantically coherent text spans, each conveying a single assessable unit of meaning. To address this task, we introduce SPANUQ, a lightweight probe that distills the uncertainty knowledge from expensive multi-sample inference into a single forward pass over LLM hidden states. SPANUQ employs a DETR-style span decoder to simultaneously detect spans and estimate their uncertainty via a Mixture of Beta distribution, trained with a principled combination of Beta NLL regression and contrastive ranking objectives. We construct SPANUQ-BENCH, the first span-level uncertainty benchmark comprising 20K prompts, 293K annotated spans, and continuous soft labels derived from multi-sample claim verification. Experiments on five LLM backbones show that SPANUQ consistently achieves the best span-level uncertainty quality , outperforming the strongest probe baseline and all sampling-based methods while being 10~20x faster. Its DETR-based span detector attains 0.910 F1, surpassing the best heuristic by 39.4%, enabling precise error localization that sequence-level methods cannot provide. The same architecture ports to five LLMs spanning two model families, with one probe trained per backbone, and we additionally observe that sequence-level uncertainty is partially decomposable, suggesting that span-level estimation subsumes sequence-level as a special case. The project page is available this http URL.
| Comments: | Accepted by NeurIPS 2026. The project page is available at this https URL |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2607.05721 [cs.CL] |
| (or arXiv:2607.05721v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2607.05721 arXiv-issued DOI via DataCite |
Submission history
From: Yimeng Zhang [view email]
[v1]
Tue, 7 Jul 2026 01:09:46 UTC (57 KB)
[v2]
Tue, 29 Sep 2026 19:55:12 UTC (67 KB)
[v3]
Wed, 4 Nov 2026 07:01:50 UTC (68 KB)
来源：arXiv cs.CL · arxiv.org