{
    "Name": "HalluVerse25",
    "Dialect Subsets": [],
    "HF Link": "https://huggingface.co/datasets/sabdalja/HalluVerse3",
    "Link": "https://huggingface.co/datasets/sabdalja/HalluVerse3",
    "License": "CC BY-NC 4.0",
    "Year": 2025,
    "Language": "multilingual",
    "Dialect": "Modern Standard Arabic",
    "Source": [
        "wikipedia"
    ],
    "Domain": [
        "politics",
        "culture",
        "science",
        "art"
    ],
    "Form": "text",
    "Annotation Style": [
        "LLM annotation",
        "human validation",
        "human annotation"
    ],
    "Description": "Fine-grained multilingual LLM hallucination dataset for English, Arabic, and Turkish.",
    "Volume": 828.0,
    "Unit": "sentences",
    "Ethical Risks": "Low",
    "Provider": [
        "Hamad Bin Khalifa University"
    ],
    "Derived From": [],
    "Paper Title": "HalluVerse25: Fine-grained Multilingual Benchmark Dataset for LLM Hallucinations",
    "Paper Link": "https://arxiv.org/pdf/2503.07833.pdf",
    "Script": "Arab",
    "Tokenized": false,
    "Host": "HuggingFace",
    "Access": "Upon-Request",
    "Cost": "",
    "Has Splits": false,
    "Partial": false,
    "Tasks": [
        "hallucination detection"
    ],
    "Venue Title": "arXiv",
    "Venue Type": "preprint",
    "Venue Name": "arXiv",
    "Authors": [
        "Samir Abdaljalil",
        "Hasan Kurban",
        "Erchin Serpedin"
    ],
    "Affiliations": [
        "Texas A&M University",
        "Hamad Bin Khalifa University"
    ],
    "Abstract": "Large Language Models (LLMs) are increasingly used in various contexts, yet remain prone to generating non-factual content, commonly referred to as \u201challucinations\u201d. The literature categorizes hallucinations into several types, including entity-level, relation-level, and sentence-level hallucinations. However, existing hallucination datasets often fail to capture fine-grained hallucinations in multilingual settings. In this work, we introduce HalluVerse25, a multilingual LLM hallucination dataset that categorizes fine-grained hallucinations in English, Arabic, and Turkish. Our dataset construction pipeline uses an LLM to inject hallucinations into factual biographical sentences, followed by a rigorous human annotation process to ensure data quality. We evaluate several LLMs on HalluVerse25, providing valuable insights into how proprietary models perform in detecting LLM-generated hallucinations across different contexts.",
    "Added By": "Zaid Alyafeai"
}