Skip to main navigation Skip to search Skip to main content

UHDF: Hallucination Detection Using Open Source Models Beyond Close Source Models Methods

  • Dongxu Liu
  • , Bufan Xu
  • , Zhilong Zhao
  • , Bing Xu*
  • , Muyun Yang
  • *Corresponding author for this work
  • Harbin Institute of Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

With the emergence of multimodal large models, the problem of hallucination has been plaguing their development and deployment. How to reliably detect the presence of hallucinations in mLLMs has become an important issue. We propose UHDF, which replaces the closed-source models in it with open-source models by improving UniHD [15], and dramatically outperforms it. By optimizing the external information used in UniHD and achieving decoupling between different external information sources, we minimize the hallucinations introduced in pipeline, and thus improve the effectiveness of hallucinations detection. UHDF using the open-source model outperforms UniHD using the closed-source model (GPT-4v), achieving 86.6% (dev set)/85.3% (test set) on MacroF1 and achieved the first place in NLPCC2024 Shared Task 10 Track1 (Open Source). Our code and models are available at https://github.com/codetalker125/UHDF.

Original languageEnglish
Title of host publicationNatural Language Processing and Chinese Computing - 13th National CCF Conference, NLPCC 2024, Proceedings
EditorsDerek F. Wong, Zhongyu Wei, Muyun Yang
PublisherSpringer Science and Business Media Deutschland GmbH
Pages389-399
Number of pages11
ISBN (Print)9789819794423
DOIs
StatePublished - 2025
Event13th CCF International Conference on Natural Language Processing and Chinese Computing, NLPCC 2024 - Hangzhou, China
Duration: 1 Nov 20243 Nov 2024

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume15363 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference13th CCF International Conference on Natural Language Processing and Chinese Computing, NLPCC 2024
Country/TerritoryChina
CityHangzhou
Period1/11/243/11/24

Keywords

  • external information
  • hallucination detection
  • mLLM

Fingerprint

Dive into the research topics of 'UHDF: Hallucination Detection Using Open Source Models Beyond Close Source Models Methods'. Together they form a unique fingerprint.

Cite this