EMNLP Findings2025
CANDY: Benchmarking LLMs' Limitations and Assistive Potential in Chinese Misinformation Fact-Checking
Ruiling Guo, Xinwei Yang, Chen Huang, Tong Zhang, Yong Hu
Abstract
The effectiveness of large language models (LLMs) to fact-check misinformation remains uncertain, despite their growing use. To this end, we present CANDY, a benchmark designed to systematically evaluate the capabilities and limitations of LLMs in fact-checking Chinese misinformation. Specifically, we curate a carefully annotated dataset of ∼20k instances. Our analysis shows that current LLMs exhibit limitations in generating accurate fact-checking conclusions, even when enhanced with chainof-thought reasoning and few-shot prompting. To understand these limitations, we develop a taxonomy to categorize flawed LLM-generated explanations for their conclusions and identify factual fabrication as the most common failure mode. Although LLMs alone are unreliable for fact-checking, our findings indicate their considerable potential to augment human performance when deployed as assistive tools in scenarios. Our dataset and code can be accessed at https://github.com/SCUNLP/CANDY.
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
Lineage
Builds on13
- RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-Grained Correctional Human FeedbackTianyu Yu, Yuan Yao, Haoye Zhang, Taiwen He et al.CVPR 2024 · 72 citations
- A Quantitative and Qualitative Evaluation of LLM-Based Explainable Fault LocalizationSungmin Kang, Gabin An, Shin YooFSE 2024 · 69 citations
- Prompting and Evaluating Large Language Models for Proactive Dialogues: Clarification, Target-guided, and Non-collaborationYang Deng, Lizi Liao, Liang Chen, Hongru Wang et al.EMNLP Findings 2023 · 55 citations
- MCFEND: A Multi-source Benchmark Dataset for Chinese Fake News DetectionYupeng Li, Haorui He, Jin Bai, Dacheng WenWWW 2024 · 30 citations
- Missing Counter-Evidence Renders NLP Fact-Checking Unrealistic for MisinformationMax Glockner, Yufang Hou, Iryna GurevychEMNLP 2022 · 23 citations
Related papers
- Chinese SafetyQA: A Safety Short-form Factuality Benchmark for Large Language ModelsYingshui Tan, Boren Zheng, Baihui Zheng, Kerui Cao et al.ACL 2025 · 7 citations
- A Chinese Dataset for Evaluating the Safeguards in Large Language ModelsYuxia Wang, Zenan Zhai, Haonan Li, Xudong Han et al.ACL Findings 2024 · 8 citations
- Factcheck-Bench: Fine-Grained Evaluation Benchmark for Automatic Fact-checkersYuxia Wang, Revanth Gangi Reddy, Zain Muhammad Mujahid, Arnav Arora et al.EMNLP Findings 2024 · 25 citations
- CLAMBER: A Benchmark of Identifying and Clarifying Ambiguous Information Needs in Large Language ModelsTong Zhang, Peixin Qin, Yang Deng, Chen Huang et al.ACL 2024
- SafetyBench: Evaluating the Safety of Large Language ModelsZhexin Zhang, Leqi Lei, Lindong Wu, Rui Sun et al.ACL 2024
