Elamur Research⌕
☰
home / papers / 2026 / TrendFact: A Benchmark Towards Hotspot Perception in Automatic Fact-Checking

TrendFact: A Benchmark Towards Hotspot Perception in Automatic Fact-Checking

Paper · 2026 · Xiaocheng Zhang, Xi Wang, Yifei Lu et al.

Summary

Fact-checking systems built on LLMs and reasoning models (RLMs) usually spend the same effort on every claim. The authors argue this is risky. Claims that spread widely turn out to be harder to verify, and getting them wrong can cause social harm up to roughly 20 times greater. They call the missing skill Hotspot Perception Ability (HPA): spending more reasoning and evidence on high-impact claims and less on minor ones. No existing benchmark could measure it.

TrendFact fills this gap. It is a Chinese benchmark that draws claims from trending platforms and the CHEF dataset. Each claim comes with gold evidence, a human-written explanation, and popularity metadata, which together yield an influence score.

The authors add two metrics: ECS for explanation reliability and HCPI for overall HPA. Current fact-checking methods and LLM/RLM systems perform poorly on the benchmark. The proposed FactISR framework raises HPA and reduces compute. This makes TrendFact a practical testbed for impact-aware, resource-limited fact-checking.

Key points

  • Dataset size and scope: TrendFact has 7,643 samples and an evidence library of 366,634 entries. It spans five domains (public health, science, society, politics, culture) and supports three tasks: evidence retrieval, claim verification, and explanation generation.
  • Influence score: each claim's score combines four social signals (views, discussions, engagements, posts) with a GPT-assessed risk score.
  • Metrics: ECS measures whether explanations are consistent with the reasoning behind the verdict. HCPI combines ECS with the influence score to quantify HPA.
  • FactISR framework: it pairs Dynamic Evidence Augmentation (DEA), which retrieves evidence iteratively as needed, with influence-driven self-reflection (ISR), which adjusts reasoning depth. On QwQ-32B and Qwen3-32B, it raises HCPI by up to almost 4%.
  • DEA efficiency: DEA keeps accuracy steady while cutting reasoning time by up to 30% and reasoning length by up to 50%. Code and data are available at github.com/zxc123cc/TrendFact.

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

🎓Tsinghua UniversityBeijing · China415 works · 94,028 citations🎓Harbin Institute of TechnologyHarbin · China185 works · 29,140 citations🎓Beihang UniversityBeijing · China86 works · 9,097 citations🎓Northeastern UniversityBoston · United States68 works · 7,316 citations🎓East China Normal UniversityShanghai · China69 works · 2,490 citations🎓National University of Defense TechnologyChangsha · China20 works · 1,339 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksSocial Sciences · 1,954 worksSociology and Political Science · 626 works

🧭 Topics 2Research subjects and languages the work is about.Reasoning · 1,685 worksEvaluation & benchmarks · 1,341 works

Cite

Cite as Xiaocheng Zhang, Xi Wang, Yifei Lu, Jianing Wang, Zhuangzhuang Ye, Mengjiao Bao, Peng Yan and Xiaohong Su. 2026. TrendFact: A Benchmark Towards Hotspot Perception in Automatic Fact-Checking. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1219

@inproceedings{zhang2026trendfact,
  title = {{TrendFact: A Benchmark Towards Hotspot Perception in Automatic Fact-Checking}},
  author = {Xiaocheng Zhang, Xi Wang, Yifei Lu, Jianing Wang, Zhuangzhuang Ye, Mengjiao Bao, Peng Yan and Xiaohong Su},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1219},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1219}
}

References

25 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. 2024. Qwen2 technical report. find
  2. Rami Aly, Zhijiang Guo, Michael Schlichtkrull, James Thorne, Andreas Vlachos, Christos Christodoulopoulos, Oana Cocarascu, and Arpit Mittal. 2021. Feverous: Fact extraction and verification over unstructured and structured information. arXiv preprint arXiv:2106.05707. — [arXiv:2106.05707](https://arxiv.org/abs/2106.05707) find
  3. Eric Msughter Aondover, Uchendu Chinelo Ebele, Timothy Ekeledirichukwu Onyejelem, and Omo-lara Oluwabusayo Akin-Odukoya. 2024. Propagation of false information on covid-19 among nigerians on social media. LingLit Journal Scientific Journal for Linguistics and Literature, 5(3):158–172. find
  4. Pepa Atanasova. 2024. Generating fact checking explanations. In Accountable and Explainable Methods for Complex Reasoning over Text, pages 83–103. Springer. find
  5. Iman Munire Bilal, Preslav Nakov, Rob Procter, and Maria Liakata. 2024. Generating unsupervised abstractive explanations for rumour verification. arXiv preprint arXiv:2401.12713. — [arXiv:2401.12713](https://arxiv.org/abs/2401.12713) find
  6. Igor Calzada. 2022. Citizens’ data privacy in china: The state of the art of the personal information protection law (pipl). Smart Cities, 5(3):1129–1150. find
  7. Jianlv Chen, Shitao Xiao, Peitian Zhang, Kun Luo, Defu Lian, and Zheng Liu. 2024. Bge m3-embedding: Multi-lingual, multi-functionality, multi-granularity text embeddings through self-knowledge distillation. arXiv preprint arXiv:2402.03216.
  8. Jifan Chen, Aniruddh Sriram, Eunsol Choi, and Greg Durrett. 2022. Generating literal and implied sub-questions to fact-check complex claims. arXiv preprint arXiv:2205.06938. — [arXiv:2205.06938](https://arxiv.org/abs/2205.06938) find
  9. Jihong Chen and Jiabin Sun. 2021. Understanding the chinese data security law. International Cybersecurity Law Review, 2(2):209–221. find
  10. Lingjiao Chen, Matei Zaharia, and James Zou. 2023. Frugalgpt: How to use large language models while reducing cost and improving performance. arXiv preprint arXiv:2305.05176.
  11. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948.
  12. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, et al. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276.
  13. Yichen Jiang, Shikha Bordia, Zheng Zhong, Charles Dognin, Maneesh Singh, and Mohit Bansal. 2020. Hover: A dataset for many-hop fact extraction and claim verification. arXiv preprint arXiv:2011.03088.
  14. Wei-Yu Kao and An-Zi Yen. 2024. How we refute claims: Automatic fact-checking through flaw identification and explanation. In Companion Proceedings of the ACM on Web Conference 2024, pages 758–761. find
  15. Ying-Jia Lin, Chun-Yi Lin, Chia-Jen Yeh, Yi-Ting Li, Yun-Yu Hu, Chih-Hao Hsu, Mei-Feng Lee, and Hung-Yu Kao. 2024. Cfever: A chinese fact extraction and verification dataset. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 18626–18634. find
  16. Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, et al. 2024. Deepseek-v3 technical report. arXiv preprint arXiv:2412.19437. find
  17. Liangming Pan, Xiaobao Wu, Xinyuan Lu, Anh Tuan Luu, William Yang Wang, Min-Yen Kan, and Preslav Nakov. 2023. Fact-checking complex claims with program-guided reasoning. arXiv preprint arXiv:2305.12744.
  18. Kashyap Popat, Subhabrata Mukherjee, Andrew Yates, and Gerhard Weikum. 2018. Declare: Debunking fake news and false claims using evidence-aware deep learning. arXiv preprint arXiv:1809.06416.
  19. Anku Rani, SM Tonmoy, Dwip Dalal, Shreya Gautam, Megha Chakraborty, Aman Chadha, Amit Sheth, and Amitava Das. 2023. Factify-5wqa: 5w aspect-based fact verification through question answering. arXiv preprint arXiv:2305.04329. find
  20. Michael Schlichtkrull, Zhijiang Guo, and Andreas Vlachos. 2024. Averitec: A dataset for real-world claim verification with evidence from the web. Advances in Neural Information Processing Systems, 36.
  21. Connie Moon Sehat, Ryan Li, Peipei Nie, Tarunima Prabhakar, and Amy X Zhang. 2024. Misinformation as a harm: structured approaches for fact-checking prioritization. Proceedings of the ACM on human-computer interaction, 8(CSCW1):1–36. find
  22. Kirill Solovev and Nicolas Pröllochs. 2022. Moral emotions shape the virality of covid-19 misinformation on social media. In Proceedings of the ACM web conference 2022, pages 3706–3717. find
  23. James Thorne and Andreas Vlachos. 2017. An extensible framework for verification of numerical claims. In Proceedings of the Software Demonstrations of the 15th Conference of the European Chapter of the Association for Computational Linguistics, pages 37–40. Association for Computational Linguistics. find
  24. James Thorne, Andreas Vlachos, Christos Christodoulopoulos, and Arpit Mittal. 2018. Fever: a large-scale dataset for fact extraction and verification. arXiv preprint arXiv:1803.05355.
  25. Category F-D F-C T-CD T-PC T-FC Acc 98.96 96.82 95.68 96.34 *Table 8: Human-LLM agreement rates (%) across five ECS categories, based on manual expert validation of over 100 randomly sampled instances.* find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhang-2026-trendfact-benchmark-hotspot-perception-automatic/