Elamur Research⌕
☰
home / papers / 2026 / Pause or Fabricate? Training Language Models for Grounded Reasoning

Pause or Fabricate? Training Language Models for Grounded Reasoning

Paper · 2026 · Yiwen Qiu, Linjuan Wu, Yizhou Liu et al.

Summary

The paper examines a failure mode the authors call ungrounded reasoning. When a question lacks a needed fact, a language model often notices the gap early but then invents a value and reasons confidently to a wrong answer. The authors argue that the cause is weak awareness of when valid inference is possible, not weak reasoning skill. Their method, GRIL (Grounded Reasoning via Interactive Reinforcement Learning), is a multi-turn RL framework with two stages:

  • Clarify and pause: the model checks whether the input is sufficient. Spotting a missing premise earns a reward that shrinks the later it happens.
  • Grounded reasoning: after detection, a simulated environment supplies the missing fact, and the model is rewarded only for a correct final answer.

Fabricated answers and needless clarification requests are both penalized. Training and evaluation use math benchmarks with premises deliberately removed (GSM8K-Insufficient, MetaMATH-Insufficient) and Qwen2.5 models. GRIL sharply improves gap detection and task success and shortens responses. It does not hurt accuracy on complete problems, transfers to non-math tasks, and copes with unhelpful user replies. This matters for real dialogues, where users often omit key details.

Key points

  • Diagnostic metrics: GapRatio is the share of tokens produced after the first sign of uncertainty. Premise Detection Rate is how often the model flags an input as incomplete. In preliminary tests, base models showed GapRatios above 47.7% and detection rates below 41%.
  • Main in-domain result: on GSM8K-Insufficient with Qwen2.5-1.5B, Premise Detection rose from 4.6% to 90.8% and Success Rate from 1.8% to 61.6%. Response length fell by over 40%, and standard GSM8K accuracy improved by up to 7%.
  • Where the gains come from: detection F1 scores of 0.908 and 0.931 indicate a precise decision boundary. For the 1.5B model, success after a correct detection (DCR) rose from 37.3% to 66.9%. Success after forced feedback without detection (NCR) rose from 36.9% to 48.2%.
  • Out-of-domain transfer: gains carried over to modified HotpotQA and CommonsenseQA with premises removed. The 1.5B model gained more than 18 points in Success Rate on CQA.
  • Robustness to imperfect users: for the 1.5B model, success with noisy, off-topic feedback rose from 12.8% to 47.2%. With evasive replies such as "I don't know," it rose from 62.5% to 86.7%.

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

🏢TencentShenzhen · China134 works · 19,743 citations🎓Zhejiang UniversityHangzhou · China172 works · 12,988 citations🏢Xiaohongshu (Little Red Book)Shanghai · China11 works · 107 citations

 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 worksComputer Vision and Pattern Recognition · 2,869 works

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

Cite

Cite as Yiwen Qiu, Linjuan Wu, Yizhou Liu, Yuchen Yan, Jin Ma, Xu Tan, Yao Hu, Daoxin Zhang, Wenqi Zhang, Weiming Lu, Jun Xiao and Yongliang Shen. 2026. Pause or Fabricate? Training Language Models for Grounded Reasoning. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.1155

@inproceedings{qiu2026pause,
  title = {{Pause or Fabricate? Training Language Models for Grounded Reasoning}},
  author = {Yiwen Qiu, Linjuan Wu, Yizhou Liu, Yuchen Yan, Jin Ma, Xu Tan, Yao Hu, Daoxin Zhang, Wenqi Zhang, Weiming Lu, Jun Xiao and Yongliang Shen},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1155},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1155}
}

References

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

  1. Simran Arora, Patrick S. H. Lewis, Angela Fan, Jacob Kahn, and Christopher Ré. 2023. Reasoning over public and private data in retrieval-based systems. Trans. Assoc. Comput. Linguistics, 11:902–921. find
  2. Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, Nicholas Joseph, Saurav Kadavath, Jackson Kernion, Tom Conerly, Sheer El Showk, Nelson Elhage, Zac Hatfield-Dodds, Danny Hernandez, Tristan Hume, and 12 others. 202… find
  3. Xiang Hu, Zujie Wen, Yafang Wang, Xiaolong Li, and Gerard de Melo. 2020. Interactive question clarification in dialogue via reinforcement learning. In Proceedings of the 28th International Conference on Computational Linguistics, COLING 2020 - Industry Track, Online, December 12, 2020, pages 78–89.… find
  4. Maciej Besta, Nils Blach, Ales Kubicek, Robert Gerstenberger, Michal Podstawski, Lukas Gianinazzi, Joanna Gajda, Tomasz Lehmann, Hubert Niewiadomski, Piotr Nyczyk, and Torsten Hoefler. 2024. Graph of thoughts: Solving elaborate problems with large language models. In Thirty-Eighth AAAI Conference on… find
  5. Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qianglong Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, and Ting Liu. 2025. A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. ACM Trans. Inf. Syst., 43(2):42:1–42:55.
  6. Yuhua Jiang, Yuwen Xiong, Yufeng Yuan, Chao Xin, Wenyuan Xu, Yu Yue, Qianchuan Zhao, and Lin Yan. 2025. Pag: Multi-turn reinforced llm self-correction with policy as generative verifier. Preprint, arXiv:2506.10406.
  7. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. Training verifiers to solve math word problems. CoRR, abs/2110.14168.
  8. Bowen Jin, Hansi Zeng, Zhenrui Yue, Jinsung Yoon, Sercan Arik, Dong Wang, Hamed Zamani, and Jiawei Han. 2025. Search-r1: Training llms to reason and leverage search engines with reinforcement learning. arXiv preprint arXiv:2503.09516.
  9. DeepSeek-AI, Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, Xiaokang Zhang, Xingkai Yu, Yu Wu, Z. F. Wu, Zhibin Gou, Zhihong Shao, Zhuoshu Li, Ziyi Gao, and 181 others. 2025. Deepseek-r1: Incentivizing reasoning capability in ll…
  10. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  11. Chenrui Fan, Ming Li, Lichao Sun, and Tianyi Zhou. 2025. Missing premise exacerbates overthinking: Are reasoning models losing critical thinking skill? CoRR, abs/2504.06514. — [arXiv:2504.06514](https://arxiv.org/abs/2504.06514) find
  12. Yichao Fu, Xuewei Wang, Yuandong Tian, and Jiawei Zhao. 2025. Deep think with confidence. CoRR, abs/2508.15260. find
  13. Yujian Gan, Changling Li, Jinxia Xie, Luou Wen, Matthew Purver, and Massimo Poesio. 2024. Clarq-llm: A benchmark for models clarifying and requesting information in task-oriented dialog. arXiv preprint arXiv:2409.06097. — [arXiv:2409.06097](https://arxiv.org/abs/2409.06097) find
  14. Zhenyu Gao. 2025. Modeling reasoning as markov decision processes: A theoretical investigation into nlp transformer models. find
  15. Qianyue Hao, Sibo Li, Jian Yuan, and Yong Li. 2025. Rl of thoughts: Navigating llm reasoning with inference-time reinforcement learning. arXiv preprint arXiv:2505.14140. — [arXiv:2505.14140](https://arxiv.org/abs/2505.14140) find
  16. Saurav Kadavath, Tom Conerly, Amanda Askell, Tom Henighan, Dawn Drain, Ethan Perez, Nicholas Schiefer, Zac Hatfield-Dodds, Nova DasSarma, Eli Tran-Johnson, and 1 others. 2022. Language models (mostly) know what they know. arXiv preprint arXiv:2207.05221. find
  17. Muhammad Khalifa, Rishabh Agarwal, Lajanugen Logeswaran, Jaekyeom Kim, Hao Peng, Moontae Lee, Honglak Lee, and Lu Wang. 2025. Process reward models that think. arXiv preprint arXiv:2504.16828. — [arXiv:2504.16828](https://arxiv.org/abs/2504.16828) find
  18. Lorenz Kuhn, Yarin Gal, and Sebastian Farquhar. 2022. Clam: Selective clarification for ambiguous questions with generative language models. arXiv preprint arXiv:2212.07769. find
  19. Philippe Laban, Hiroaki Hayashi, Yingbo Zhou, and Jennifer Neville. 2025. Llms get lost in multi-turn conversation. CoRR, abs/2505.06120.
  20. Kenneth Li, Oam Patel, Fernanda Viégas, Hanspeter Pfister, and Martin Wattenberg. 2023. Inference-time intervention: Eliciting truthful answers from a language model. Advances in Neural Information Processing Systems, 36:41451–41530.
  21. Xiaoxi Li, Guanting Dong, Jiajie Jin, Yuyao Zhang, Yujia Zhou, Yutao Zhu, Peitian Zhang, and Zhicheng Dou. 2025. Search-o1: Agentic search-enhanced large reasoning models. arXiv preprint arXiv:2501.05366.
  22. Hunter Lightman, Vineet Kosaraju, Yuri Burda, Harrison Edwards, Bowen Baker, Teddy Lee, Jan Leike, John Schulman, Ilya Sutskever, and Karl Cobbe. 2023. Let’s verify step by step. In The Twelfth International Conference on Learning Representations.
  23. Yuxin Liu, Chaojie Gu, Yihang Zhang, Bin Qian, and Shibo He. 2025. Reverse thinking enhances missing information detection in large language models. Preprint, arXiv:2512.10273. — [arXiv:2512.10273](https://arxiv.org/abs/2512.10273) find
  24. Liangchen Luo, Yinxiao Liu, Rosanne Liu, Samrat Phatale, Harsh Lara, Yunxuan Li, Lei Shu, Yun Zhu, Lei Meng, Jiao Sun, and Abhinav Rastogi. 2024. Improve mathematical reasoning in language models by automated process supervision. CoRR, abs/2406.06592.
  25. Sichun Luo, Yi Huang, Mukai Li, Shichang Meng, Fengyuan Liu, Zefa Hu, Junlan Feng, and Qi Liu. 2025. Clarifymt-bench: Benchmarking and improving multi-turn clarification for conversational large language models. Preprint, arXiv:2512.21120. — [arXiv:2512.21120](https://arxiv.org/abs/2512.21120) find
  26. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, Shashank Gupta, Bodhisattwa Prasad Majumder, Katherine Hermann, Sean Welleck, Amir Yazdanbakhsh, and Peter Clark. 2023. Self-refine: Iterative refinement with… find
  27. GPT-4 technical report. CoRR, Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, and 1 others. 2022. Training language models to follow instructions with human feedback. Advances in neural information process…
  28. Mathieu Rita, Florian Strub, Rahma Chaabouni, Paul Michel, Emmanuel Dupoux, and Olivier Pietquin. 2024. Countering reward over-optimization in llm with demonstration-guided reinforcement learning. arXiv preprint arXiv:2404.19409. find
  29. Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, Maria Lomeli, Eric Hambro, Luke Zettlemoyer, Nicola Cancedda, and Thomas Scialom. 2023. Toolformer: Language models can teach themselves to use tools. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural I…
  30. John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. CoRR, abs/1707.06347. find
  31. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Mingchuan Zhang, Y. K. Li, Y. Wu, and Daya Guo. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. CoRR, abs/2402.03300.
  32. Mrinank Sharma, Meg Tong, Tomasz Korbak, David Duvenaud, Amanda Askell, Samuel R Bowman, Newton Cheng, Esin Durmus, Zac Hatfield-Dodds, Scott R Johnston, and 1 others. 2023. Towards understanding sycophancy in language models. arXiv preprint arXiv:2310.13548.
  33. Shuaijie She, Junxiao Liu, Yifeng Liu, Jiajun Chen, Xin Huang, and Shujian Huang. 2025. R-PRM: reasoning-driven process reward modeling. CoRR, abs/2503.21295.
  34. Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. 2025. Hybridflow: A flexible and efficient RLHF framework. In Proceedings of the Twentieth European Conference on Computer Systems, EuroSys 2025, Rotterdam, The Netherlands, 30 March 20…
  35. 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.
  36. Yang Sui, Yu-Neng Chuang, Guanchu Wang, Jiamu Zhang, Tianyi Zhang, Jiayi Yuan, Hongyi Liu, Andrew Wen, Shaochen Zhong, Na Zou, and 1 others. 2025. Stop overthinking: A survey on efficient reasoning for large language models. arXiv preprint arXiv:2503.16419.
  37. Peiyi Wang, Lei Li, Zhihong Shao, Runxin Xu, Damai Dai, Yifei Li, Deli Chen, Yu Wu, and Zhifang Sui. 2024. Math-shepherd: Verify and reinforce llms step-by-step without human annotations. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Paper…
  38. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc V. Le, Ed H. Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2023a. Self-consistency improves chain of thought reasoning in language models. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5…
  39. Yanbo Wang, Yongcan Yu, Jian Liang, and Ran He. 2025a. A comprehensive survey on trustworthiness in reasoning with large language models. arXiv preprint arXiv:2509.03871. find
  40. Zekun Wang, Ge Zhang, Kexin Yang, Ning Shi, Wangchunshu Zhou, Shaochun Hao, Guangzheng Xiong, Yizhi Li, Mong Yuan Sim, Xiuying Chen, and 1 others. 2023b. Interactive natural language processing. arXiv preprint arXiv:2305.13246. — [arXiv:2305.13246](https://arxiv.org/abs/2305.13246) find
  41. Zihan Wang, Kangrui Wang, Qineng Wang, Pingyue Zhang, Linjie Li, Zhengyuan Yang, Xing Jin, Kefan Yu, Minh Nhat Nguyen, Licheng Liu, Eli Gottlieb, Yiping Lu, Kyunghyun Cho, Jiajun Wu, Li Fei-Fei, Lijuan Wang, Yejin Choi, and Manling Li. 2025b. RAGEN: understanding self-evolution in LLM agents via mul… find
  42. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Brian Ichter, Fei Xia, Ed H. Chi, Quoc V. Le, Yuchen Yan, Yongliang Shen, Yang Liu, Jin Jiang, Xin Xu, Mengdi Zhang, Jian Shao, and Yueting Zhuang. 2025. Mathfimer: Enhancing mathematical reasoning by expanding reasoning steps through fill-in-t… find
  43. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 40 others. 2025. Qwen3 technical report. CoRR, abs/2505.09388.
  44. 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, Junyang Lin, Kai Dang, and 22 others. 2024. Qwen2.5 technical report. CoRR, abs/2412.1511… find
  45. Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William W. 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, Brussel…
  46. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. 2023a. Tree of thoughts: Deliberate problem solving with large language models. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023,… find
  47. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik R. Narasimhan, and Yuan Cao. 2023b. React: Synergizing reasoning and acting in language models. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net.
  48. Longhui Yu, Weisen Jiang, Han Shi, Jincheng Yu, Zhengying Liu, Yu Zhang, James T. Kwok, Zhenguo Li, Adrian Weller, and Weiyang Liu. 2024. Metamath: Bootstrap your own mathematical questions for large language models. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vie…
  49. Michael JQ Zhang, W Bradley Knox, and Eunsol Choi. 2024. Modeling future conversation turns to Xuan Zhang, Yongliang Shen, Zhe Zheng, Linjuan Wu, Wenqi Zhang, Yuchen Yan, Qiuying Peng, Jun Wang, and Weiming Lu. 2025. Asktoact: Enhancing llms tool use via self-correcting clarification. CoRR, abs/2503… find
  50. Yifei Zhou, Andrea Zanette, Jiayi Pan, Sergey Levine, and Aviral Kumar. 2024. Archer: Training language model agents via hierarchical multi-turn RL. In Forty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024. Open-Review.net. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/qiu-2026-pause-or-fabricate-training-language/