Elamur Research⌕
☰
home / papers / 2026 / MFMDQwen: Multilingual Financial Misinformation Detection Based on Large Language Model

MFMDQwen: Multilingual Financial Misinformation Detection Based on Large Language Model

Paper · 2026 · Zhiwei Liu, Yuyan Wang, Yuechen Jiang et al.

Summary

False or manipulated financial content is hard to detect because it often hides in small numerical edits, exaggerated sentiment or inverted cause–effect claims. Most LLM-based detectors so far, such as FMDLlama, handle only English and a single task. The authors tackle multilingual financial misinformation detection (MFMD) by pooling nine existing datasets in English, Chinese, Greek and Bengali. These cover claim verification, fact-checking, fake-post detection and social media manipulation, and the authors convert them into one shared instruction format. The result is MFMD4Instruction for training and MFMDBench for evaluation. They then fully fine-tune Qwen3-8B on the merged multi-task data, framing detection as answer generation, and release the resulting model as MFMDQwen. The baselines are Qwen3, Qwen2.5, Llama 3.x and FMDLlama, with up to 72B parameters. MFMDQwen gets the best Macro-F1 on six of nine datasets, with its largest gains on Chinese data. General-purpose models still do better on evidence-based English verification (FinDVer). The work provides the first open multilingual model and resource for this high-risk domain.

Key points

  • Data sources: FinDVer and RFC (RFC-Bench, Yahoo Finance paragraphs with GPT-4.1 perturbations) for English; CHEF and MDFEND (Weibo posts) for Chinese; BanMANI for Bengali; and Global4Languages, with 144 aligned claims each in EN/ZH/BN/GR.
  • Dataset sizes: MFMD4Instruction has 4,837 training samples and MFMDBench has 2,432 test samples, for 7,269 in total.
  • Training setup: full-parameter SFT of Qwen3-8B (lr 1e-5, 2 epochs, batch 128, DeepSpeed ZeRO-3) on four NVIDIA L40s GPUs.
  • Strongest results: on CHEF, MFMDQwen reaches F1 0.843, against about 0.32 for the best baselines. It also scores 0.913 on MDFEND and 0.803 on GlobalBe.
  • Weaker results and failure modes: on FinDVer, MFMDQwen scores F1 0.673, below Qwen3-32B-R at 0.857. Llama3.1-8B and FMDLlama often refuse to answer, and reasoning variants sometimes "overthink" their way to wrong answers.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓University of EdinburghEdinburgh · United Kingdom359 works · 77,352 citations🎓Columbia UniversityNew York · United States241 works · 54,053 citations🎓University of ManchesterManchester · United Kingdom87 works · 44,938 citations🎓Stevens Institute of TechnologyHoboken · United States10 works · 95 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksSocial Sciences · 1,772 worksSociology and Political Science · 577 works

🧭 Topics 1Research subjects and languages the work is about.Multilingual & cross-lingual NLP · 828 works

Cite

Cite as Zhiwei Liu, Yuyan Wang, Yuechen Jiang, Yupeng Cao, Tianlei Zhu, Xiaorui Guo, Zhiyang Deng, Zhiyuan Yao, Xiao-Yang Liu, Jimin Huang and Sophia Ananiadou. 2026. MFMDQwen: Multilingual Financial Misinformation Detection Based on Large Language Model. Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.mellm-1.7

@inproceedings{liu2026mfmdqwen,
  title = {{MFMDQwen: Multilingual Financial Misinformation Detection Based on Large Language Model}},
  author = {Zhiwei Liu, Yuyan Wang, Yuechen Jiang, Yupeng Cao, Tianlei Zhu, Xiaorui Guo, Zhiyang Deng, Zhiyuan Yao, Xiao-Yang Liu, Jimin Huang and Sophia Ananiadou},
  booktitle = {Proceedings of the 1st Workshop on Multilinguality in the Era of Large Language Models (MeLLM 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.mellm-1.7},
  url = {https://doi.org/10.18653/v1/2026.mellm-1.7}
}

References

14 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Yupeng Cao, Haohang Li, Yangyang Yu, and Shashidhar Reddy Javaji. 2025. Capybara at the financial misinformation detection challenge task: chain-of-thought enhanced financial misinformation detection. In Proceedings of the Joint Workshop of the 9th Financial Technology and Natural Language Processin… find
  2. Zhiwei Liu, Yupen Cao, Yuechen Jiang, Mohsinul Kabir, Polydoros Giannouris, Chen Xu, Ziyang Xu, Tianlei Zhu, Tariquzzaman Faisal, Triantafillos Papadopoulos, et al. 2026a. Same claim, different judgment: Benchmarking scenario-induced bias in multilingual financial misinformation detection. arXiv pre…
  3. Zhiwei Liu, Runteng Guo, Baojie Qu, Yuechen Jiang, Min Peng, Qianqian Xie, and Sophia Ananiadou. 2026b. Raar: Retrieval augmented agentic reasoning for cross-domain misinformation detection. arXiv preprint arXiv:2601.04853. find
  4. Junying Chen, Zhenyang Cai, Ke Ji, Xidong Wang, Wanlong Liu, Rongsheng Wang, Jianye Hou, and Benyou Wang. 2024. Huatuogpt-o1, towards medical complex reasoning with llms. arXiv preprint arXiv:2412.18925. find
  5. Zhiwei Liu, Paul Thompson, Jiaqi Rong, and Sophia Ananiadou. 2025b. Conspemollm-v2: A robust and stable model to detect sentiment-transformed conspiracy theories. arXiv preprint arXiv:2505.14917. — [arXiv:2505.14917](https://arxiv.org/abs/2505.14917) find
  6. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  7. Zhiwei Liu, Keyi Wang, Zhuo Bao, Xin Zhang, Jiping Dong, Kailai Yang, Mohsinul Kabir, Polydoros Giannouris, Rui Xing, Park Seongchan, et al. 2025c. Finnlp-fnp-llmfinlegal-2025 shared task: financial misinformation detection challenge task. In Proceedings of the Joint Workshop of the 9th Financial Te… find
  8. Xuming Hu, Zhijiang Guo, GuanYu Wu, Aiwei Liu, Lijie Wen, and Philip S Yu. 2022. Chef: A pilot chinese dataset for evidence-based fact-checking. In Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 33… find
  9. Zhiwei Liu, Xin Zhang, Kailai Yang, Qianqian Xie, Jimin Huang, and Sophia Ananiadou. 2025d. Fmdllama: Financial misinformation detection based on large language models. In Companion Proceedings of the ACM on Web Conference 2025, pages 1153– 1157. find
  10. Qianqian Xie, Weiguang Han, Xiao Zhang, Yanzhao Lai, Min Peng, Alejandro Lopez-Lira, and Jimin Huang. 2023. Pixiu: A large language model, instruction data and evaluation benchmark for finance. arXiv preprint arXiv:2306.05443. — [arXiv:2306.05443](https://arxiv.org/abs/2306.05443) find
  11. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  12. Zheyang Luo, Guangbin Zhang, Jiahao Xiao, Xuankang Zhang, Yulin Dou, and Jiangming Liu. 2025. Fmd-mllama at the financial misinformation detection challenge task: multimodal reasoning and evidence generation. In Proceedings of the Joint Workshop of the 9th Financial Technology and Natural Language P… find
  13. Yilun Zhao, Yitao Long, Tintin Jiang, Chengye Wang, Weiyuan Chen, Hongjun Liu, Xiangru Tang, Yiming Zhang, Chen Zhao, and Arman Cohan. 2024. Findver: Explainable claim verification over long and hybrid-content financial documents. In Proceedings of the 2024 Conference on Empirical Methods in Natural… find
  14. Shakalya Nag. 2025. Financial misinformation and trading manipulation using large language models (llms). In International IOT, Electronics and Mechatronics Conference, pages 1–15. Springer. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/liu-2026-mfmdqwen-multilingual-financial-misinformation-detection/