---
title: "HalluTruthQA-4K：阿拉伯语幻觉检测与事实核验的细粒度语料与标注流程"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.03966v1"
permalink: "https://feed.astralor.com/items/itm_52a157b776cbd62b656db5d8"
published: "2026-08-04T17:33:38.000Z"
score: 80
channel: "ai"
tags: ["arXiv","HalluTruthQA-4K","事实核验","幻觉检测","数据集构建","阿拉伯语"]
---

# HalluTruthQA-4K：阿拉伯语幻觉检测与事实核验的细粒度语料与标注流程

## AI 点评

> 这份工作最值得关注的地方在于把幻觉标注从整体标签细化到了字符级跨度，并针对阿拉伯语这种缺乏评测资源的语言。对做多语言NLP评测或关注阿拉伯语大模型可靠性的研究者来说，这是一个可以直接拿来用的资源。不过目前只是预印本，后续需要观察数据质量和共享任务结果是否经得起检验。

## 速览

大语言模型在生成流利阿拉伯语回答时，常引入难以识别和核验的事实错误。现有阿拉伯语幻觉资源通常只对整段回答做二元标注，缺乏对具体错误内容、错误原因及正确事实答案的细粒度信息。本文提出HalluTruthQA-4K，将原始语料扩展至4000条专家精选问答实例，覆盖伊斯兰知识、历史、科学和地理四个知识密集型领域。每条实例包含阿拉伯语问题、模型生成回答、经核实的参考答案及五个合理干扰项。幻觉回答额外标注了字符级错误跨度、人工解释和分层幻觉类型。该语料包含1643条幻觉回答和2357条非幻觉回答，共标注1843个错误跨度，并详细描述了资源构建、标注指南、分类体系、数据格式及标注一致性，为阿拉伯语模型的幻觉检测、错误定位、解释生成和事实核验提供可复用资源。

## 关键点

- HalluTruthQA-4K包含4000条专家精选阿拉伯语问答实例，覆盖伊斯兰知识、历史、科学和地理四个领域。
- 相比仅做整句二元标注的现有资源，该语料提供字符级错误跨度、人工解释和分层幻觉类型。
- 语料含1643条幻觉回答和2357条非幻觉回答，共标注1843个错误跨度。
- 作为HalluScoring 2026共享任务Track 2官方数据集，可用于幻觉检测、错误定位、解释生成和事实核验。

## 全文

Large language models can generate fluent Arabic answers while introducing factual errors that are difficult to identify and verify. Existing Arabic hallucination resources often assign a binary label to an entire response, indicating whether it is hallucinated or non-hallucinated, but provide limited information about the exact erroneous content, the reason for the error, or the correct factual answer. We present HalluTruthQA-4K, an expanded version of the HalluTruthQA resource containing 4,000 expert-curated Arabic question-answering instances across four knowledge-intensive domains: Islamic knowledge, history, science, and geography. Serving as the official dataset for Track 2 of the HalluScoring 2026 shared task, HalluTruthQA-4K extends our original corpus to 4,000 instances. Each instance pairs an Arabic question with a model-generated response, a verified reference answer, and five plausible distractors. Hallucinated responses are additionally annotated with character-level erroneous spans, human-written explanations, and hierarchical hallucination types. The corpus contains 1,643 hallucinated and 2,357 non-hallucinated responses, with 1,843 annotated erroneous spans. We describe the resource construction and annotation methodology, including question selection, controlled answer generation, candidate construction, expert annotation, independent verification, adjudication, and quality control. We also document the annotation guidelines, taxonomy, data format, inter-annotator agreement, and corpus statistics. HalluTruthQA-4K provides a reusable resource for hallucination detection, span-level error localization, explanation generation, factual verification, and the broader evaluation of factual reliability in Arabic language models.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_52a157b776cbd62b656db5d8