ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2403.04307
26
14

HaluEval-Wild: Evaluating Hallucinations of Language Models in the Wild

7 March 2024
Zhiying Zhu
Yiming Yang
Zhiqing Sun
    HILM
    VLM
ArXivPDFHTML
Abstract

Hallucinations pose a significant challenge to the reliability of large language models (LLMs) in critical domains. Recent benchmarks designed to assess LLM hallucinations within conventional NLP tasks, such as knowledge-intensive question answering (QA) and summarization, are insufficient for capturing the complexities of user-LLM interactions in dynamic, real-world settings. To address this gap, we introduce HaluEval-Wild, the first benchmark specifically designed to evaluate LLM hallucinations in the wild. We meticulously collect challenging (adversarially filtered by Alpaca) user queries from existing real-world user-LLM interaction datasets, including ShareGPT, to evaluate the hallucination rates of various LLMs. Upon analyzing the collected queries, we categorize them into five distinct types, which enables a fine-grained analysis of the types of hallucinations LLMs exhibit, and synthesize the reference answers with the powerful GPT-4 model and retrieval-augmented generation (RAG). Our benchmark offers a novel approach towards enhancing our comprehension and improvement of LLM reliability in scenarios reflective of real-world interactions. Our benchmark is available at https://github.com/Dianezzy/HaluEval-Wild.

View on arXiv
Comments on this paper