无需微调，跨语言通用的模型置信度估计方法。
Shared Doubt: Zero-Shot Cross-Lingual Confidence Estimation for Language Models
- 用轻量线性探测器从中间层表征直接预测答案正确性。
- 在未见过的多种语言上零样本泛化，性能优于现有方法。
- 发现中层特征蕴含共享置信信号，适合多语言部署场景。
置信度估计（CE）旨在量化大语言模型预测的可靠性，但现有研究多集中于英语，忽视了多语言使用现实。许多方法在跨语言迁移时性能下降或需重新训练。本文探究多语言大模型是否在开放问答任务中编码了可跨语言转移的置信度特征。采用轻量级线性探测器，直接从中间表示预测答案正确性。该探测器在单语训练后，可在无目标语言标注的情况下零样本泛化至多种类型语言。层权重分析与消融实验表明，置信度特征集中在各语言的中层，提示存在共享置信子空间。尽管零样本性能受源语言相似性影响，该方法仍无需微调即可提供强基线，优于其他主流置信度估计方法。
原文摘要 · Abstract (English)
Confidence estimation (CE), i.e., quantifying the reliability of a model's prediction, has attracted great interest in the context of large language models (LLMs). However, most studies focus on English, ignoring the multilingual reality of LLM usage, while many CE methods degrade or require retraining across languages. To address this gap, we investigate whether multilingual LLMs encode shared, language-transferable confidence features in open-ended question answering. We use a lightweight linear probe that predicts answer correctness directly from intermediate representations. Trained monolingually, the probe generalizes zero-shot to unseen, typologically diverse languages without target-language supervision. Learned layer weights and multiple ablations reveal that confidence features concentrate in middle layers across languages, suggesting a shared confidence subspace. While zero-shot cross-lingual performance depends on similarity to the source language, the probe provides a strong baseline without any retraining and compares favorably to other popular confidence estimation methods.
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。