Elamur Research⌕
☰
home / papers / 2026 / How Long Reasoning Chains Influence LLMs' Judgment of Answer Factuality

How Long Reasoning Chains Influence LLMs' Judgment of Answer Factuality

Paper · 2026 · Minzhu Tu, Shiyu Ni and Keping Bi

Summary

LLM judges are increasingly used in place of human annotators, but they are prone to surface biases and usually see only a final answer. This paper asks whether showing the judge the generator's reasoning trace helps or hurts its verdict on answer correctness. The authors compare two setups: the judge sees only the answer, or it sees the answer plus the reasoning chain. Generators are Qwen3 (8B, 14B, 32B) and DeepSeek-V3.1. Judges come from the Qwen3, Llama 3 and GLM-4 families, plus GPT-4o, Claude Sonnet 4.5 and DeepSeek-V3.1. Tasks cover factual QA (NQ, HotpotQA) and math (GSM8K, MATH500).

Weaker judges become much more willing to accept wrong answers once reasoning is visible. Stronger judges can sometimes use the trace to catch mistakes, yet every judge is fooled more often by DeepSeek-V3.1's polished chains. In controlled edits, the authors insert irrelevant true facts to hurt fluency or counterfactual facts to hurt factuality, and both kinds of change shift verdicts. The results matter for anyone using LLM-as-a-Judge to evaluate reasoning models.

Key points

  • Two judging conditions are compared, answer-only versus answer plus <think> reasoning, across 4 benchmarks: NQ, HotpotQA, GSM8K and MATH500.
  • Weak judges systematically inflate correctness when reasoning is shown. Strong judges (e.g., GPT-4o, Claude Sonnet 4.5) are more selective but still misled by high-quality DeepSeek-V3.1 chains.
  • Fluency was degraded by inserting correct but off-topic knowledge, and factuality by swapping in counterfactual statements. Both lowered acceptance, and the counterfactual edits had the stronger effect.
  • Edits placed at the start of a chain hurt the answer more than the same edits at the end. Under Basic-All with counterfactual content, Qwen3-8B's pass rate was 61.0% for a prefix versus 86.0% for a suffix. GPT-4o's was 18.4% versus 39.2%, compared with 48.6% on natural chains.
  • Stated limitations: the study is text-only, judges gave verdicts without generating their own reasoning, and model coverage is necessarily incomplete.

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

 Fields, topics and concepts

📚 Shelves 5The 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 worksInformation Systems · 1,437 works

🧭 Topics 2Research subjects and languages the work is about.Language models & pretraining · 3,320 worksReasoning · 1,426 works

Cite

Cite as Minzhu Tu, Shiyu Ni and Keping Bi. 2026. How Long Reasoning Chains Influence LLMs' Judgment of Answer Factuality. 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.2082

@inproceedings{tu2026how,
  title = {{How Long Reasoning Chains Influence LLMs' Judgment of Answer Factuality}},
  author = {Minzhu Tu, Shiyu Ni and Keping Bi},
  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.2082},
  url = {https://doi.org/10.18653/v1/2026.acl-long.2082}
}

References

45 references, 27 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. AI@Meta. 2024. Llama 3.1 model card. Technical report, Meta AI. Online documentation, accessed 2026-01-05. find
  2. Anthropic. 2025. Claude opus 4.5 system card. Technical report, Anthropic. Accessed: 2026-01-05. find
  3. Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadomski, Piotr Nyczyk, and 1 others. 2024. Graph of thoughts: Solving elaborate problems with large language models. In Proceedings of the AAAI conference on a… find
  4. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, and 1 others. 2020. Language models are few-shot learners. Advances in neural information processing systems, 33:1877–1901.
  5. Chi-Min Chan, Weize Chen, Yusheng Su, Jianxuan Yu, Wei Xue, Shanghang Zhang, Jie Fu, and Zhiyuan Liu. 2023. Chateval: Towards better llm-based evaluators through multi-agent debate. arXiv preprint arXiv:2308.07201.
  6. Hongyu Chen and Seraphina Goldfarb-Tarrant. 2025. Safer or luckier? llms as safety evaluators are not robust to artifacts. arXiv preprint arXiv:2503.09347.
  7. Cheng-Han Chiang and Hung-yi Lee. 2023. Can large language models be an alternative to human evaluations? In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15607–15631, Toronto, Canada. Association for Computational Linguistics. find
  8. Wei-Lin Chiang, Lianmin Zheng, Ying Sheng, Anastasios Nikolas Angelopoulos, Tianle Li, Dacheng Li, Banghua Zhu, Hao Zhang, Michael Jordan, Joseph E Gonzalez, and 1 others. 2024. Chatbot arena: An open platform for evaluating llms by human preference. In Forty-first International Conference on Machin…
  9. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, and 1 others. 2021. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168.
  10. DeepSeek-AI, Aixin Liu, Bei Feng, Bing Xue, Bingxuan Wang, Bochao Wu, Chengda Lu, Chenggang Zhao, Chengqi Deng, Chenyu Zhang, Chong Ruan, Damai Dai, Daya Guo, Dejian Yang, Deli Chen, Dongjie Ji, Erhang Li, Fangyun Lin, Fucong Dai, and 181 others. 2025. Deepseek-v3 technical report. Preprint, arXiv:2… find
  11. Yilun Du, Shuang Li, Antonio Torralba, Joshua B Tenenbaum, and Igor Mordatch. 2024. Improving factuality and reasoning in language models through multiagent debate. In Forty-first international conference on machine learning.
  12. Yann Dubois, Xuechen Li, Rohan Taori, Tianyi Zhang, Ishaan Gulrajani, Jimmy Ba, Carlos Guestrin, Percy Liang, and Tatsunori Hashimoto. 2023. Alpacafarm: A simulation framework for methods that learn from human feedback. In Thirty-seventh Conference on Neural Information Processing Systems.
  13. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021. Measuring mathematical problem solving with the math dataset. arXiv preprint arXiv:2103.03874.
  14. Fan Huang, Haewoon Kwak, Kunwoo Park, and Jisun An. 2024. Chatgpt rates natural language explanation quality like humans: But on which scales? In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 3111– 313… find
  15. Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, and 1 others. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276.
  16. Seungone Kim, Jamin Shin, Yejin Cho, Joel Jang, Shayne Longpre, Hwaran Lee, Sangdoo Yun, Seongjin Shin, Sungdong Kim, James Thorne, and 1 others. 2023. Prometheus: Inducing fine-grained evaluation capability in language models. In The Twelfth International Conference on Learning Representations.
  17. Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. 2022. Large language models are zero-shot reasoners. Advances in neural information processing systems, 35:22199– 22213.
  18. Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, and 1 others. 2019. Natural questions: a benchmark for question answering research. Transactions of the Association for Computational Lin…
  19. Junlong Li, Shichao Sun, Weizhe Yuan, Run-Ze Fan, Hai Zhao, and Pengfei Liu. 2023. Generative judge for evaluating alignment. arXiv preprint arXiv:2310.05470.
  20. Yang Liu, Dan Iter, Yichong Xu, Shuohang Wang, Ruochen Xu, and Chenguang Zhu. 2023. G-eval: NLG evaluation using gpt-4 with better human alignment. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 2511–2522, Singapore. Association for Computational Lin…
  21. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, and 1 others. 2023. Self-refine: Iterative refinement with self-feedback. Advances in neural information processing systems, 36:46534–46594. find
  22. Oscar Mañas, Benno Krojer, and Aishwarya Agrawal. 2024. Improving automatic vqa evaluation using large language models. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 4171–4179. find
  23. Arash Marioriyad, Mohammad Hossein Rohban, and Mahdieh Soleymani Baghshah. 2025. The silent judge: Unacknowledged shortcut bias in llm-as-a-judge. arXiv preprint arXiv:2509.26072. — [arXiv:2509.26072](https://arxiv.org/abs/2509.26072) find
  24. Shiyu Ni, Keping Bi, Jiafeng Guo, and Xueqi Cheng. 2024. When do llms need retrieval augmentation? mitigating llms’ overconfidence helps retrieval augmentation. In Findings of the Association for Computational Linguistics: ACL 2024, pages 11375–11388. find
  25. Shiyu Ni, Keping Bi, Jiafeng Guo, Lulu Yu, Baolong Bi, and Xueqi Cheng. 2025. Towards fully exploiting llm internal states to enhance knowledge boundary perception. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 24315–24329.
  26. Maja Pavlovic and Massimo Poesio. 2024. The effectiveness of llms as annotators: A comparative overview and empirical analysis of direct representation. In Proceedings of the 3rd Workshop on Perspectivist Approaches to NLP (NLPerspectives)@ LREC-COLING 2024, pages 100–110.
  27. Ofir Press, Muru Zhang, Sewon Min, Ludwig Schmidt, Noah A Smith, and Mike Lewis. 2023. Measuring and narrowing the compositionality gap in language models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 5687–5711. find
  28. Qwen, :, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, and 25 others. 2025. Qwen2.5 technical report. Preprint, arXiv:2412.15115. find
  29. Zhengliang Shi, Yiqun Chen, Haitao Li, Weiwei Sun, Shiyu Ni, Yougang Lyu, Run-Ze Fan, Bowen Jin, Yixuan Weng, Minjun Zhu, and 1 others. 2025. Deep research: A systematic survey. arXiv preprint arXiv:2512.02038. — [arXiv:2512.02038](https://arxiv.org/abs/2512.02038) find
  30. Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. 2023. Reflexion: Language agents with verbal reinforcement learning. Advances in neural information processing systems, 36:8634–8652.
  31. Sijun Tan, Siyuan Zhuang, Kyle Montgomery, William Yuan Tang, Alejandro Cuadron, Chenguang Wang, Raluca Popa, and Ion Stoica. 2025. Judgebench: A benchmark for evaluating LLM-based judges. In The Thirteenth International Conference on Learning Representations. find
  32. Pat Verga, Sebastian Hofstatter, Sophia Althammer, Yixuan Su, Aleksandra Piktus, Arkady Arkhangorodsky, Minjie Xu, Naomi White, and Patrick Lewis. 2024. Replacing judges with juries: Evaluating llm generations with a panel of diverse models. arXiv preprint arXiv:2404.18796.
  33. Peiyi Wang, Lei Li, Liang Chen, Zefan Cai, Dawei Zhu, Binghuai Lin, Yunbo Cao, Lingpeng Kong, Qi Liu, Tianyu Liu, and 1 others. 2024. Large language models are not fair evaluators. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pag… find
  34. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2022. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171.
  35. Yidong Wang, Zhuohao Yu, Zhengran Zeng, Linyi Yang, Cunxiang Wang, Hao Chen, Chaoya Jiang, Rui Xie, Jindong Wang, Xing Xie, and 1 others. 2023. Pandalm: An automatic evaluation benchmark for llm instruction tuning optimization. arXiv preprint arXiv:2306.05087.
  36. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824– 24837.
  37. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  38. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, and Christopher D Manning. 2018. Hotpotqa: A dataset for diverse, explainable multi-hop question answering. In Proceedings of the 2018 conference on empirical methods in natural language processing, pages 2369–…
  39. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. Advances in neural information processing systems, 36:11809–11822. find
  40. Eric Zelikman, Yuhuai Wu, Jesse Mu, and Noah Goodman. 2022. Star: Bootstrapping reasoning with reasoning. Advances in Neural Information Processing Systems, 35:15476–15488.
  41. Anqi Zhang, Yulin Chen, Jane Pan, Chen Zhao, Aurojit Panda, Jinyang Li, and He He. 2025. Reasoning models know when they’re right: Probing hidden states for self-verification. arXiv preprint arXiv:2504.05419. find
  42. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2023. Judging LLM-as-a-judge with MT-bench and chatbot arena. In Thirty-seventh Conference on Neural Information Processing…
  43. Zhipu AI. 2025. Glm-4-32b model card. https://huggingface.co/zai-org/GLM-4-32B-0414. ↗
  44. Denny Zhou, Nathanael Schärli, Le Hou, Jason Wei, Nathan Scales, Xuezhi Wang, Dale Schuurmans, Claire Cui, Olivier Bousquet, Quoc Le, and 1 others. 2022. Least-to-most prompting enables complex reasoning in large language models. arXiv preprint arXiv:2205.10625.
  45. Lianghui Zhu, Xinggang Wang, and Xinlong Wang. 2023. Judgelm: Fine-tuned large language models are scalable judges. arXiv preprint arXiv:2310.17631.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tu-2026-how-long-reasoning-chains-influence/