Elamur Research⌕
☰
home / papers / 2026 / Beyond Noise: Characterizing Creative Potential in Unverifiable LLM Hallucinations

Beyond Noise: Characterizing Creative Potential in Unverifiable LLM Hallucinations

Paper · 2026 · Yu Yan, Chunhong Zhang, Haiyu Zhao et al.

Summary

When LLMs are asked to invent new scientific or technical concepts, many of their outputs cannot be checked against existing evidence. Earlier work either tries to suppress such "unverifiable hallucinations" or praises them as creative, but nobody had examined what this space actually contains. The authors map outputs on two axes, novelty and verifiability. Within the novel-but-unverifiable quadrant, they separate Creative Synthesis (Region A), which is coherent and grounded, from Groundless Fabrication (Region B), which strings jargon together without a working mechanism. A plausibility threshold τ defines the boundary between the two.

They test this in a concept-creation task covering three technical domains. The task uses direct and structured generation strategies and both intra-domain and cross-domain pathways. Outputs are scored with automatic metrics and judged by human experts. A small but stable share of the outputs falls into Region A. A retrospective experiment also shows that an older model can reconstruct concepts published after its training cutoff. The results argue for filtering hallucinations selectively rather than discarding them all.

Key points

  • Scale: 32,400 generations from 150 source concepts per domain (Communication, Electromagnetics, Artificial Intelligence), plus 1,080 human judgments.
  • Region A share: about 4.7% of highly novel outputs were rated both novel and plausible. This share held under stricter thresholds and across models, domains, strategies and embedding choices.
  • Strategy differences:
    • Unconstrained Direct Generation reaches broader novelty.
    • Structured Innovation keeps outputs better grounded. It works in three stages: decomposing the concept, modifying its elements, and reconstructing a new concept.
  • Retrospective recovery experiment:
    • GPT-3.5-Turbo-0125 (cutoff September 2021) combined elements from Communication and AI concept pools into 400 candidates.
    • It recovered 17 of 20 (85%) target concepts published in 2024 or later, with GPT-4o judging consistency.
    • The other 383 candidates averaged a plausibility score of 3.23/4.
  • Code and limitations: code is available at github.com/YuLab1/llm-concept-creation. Stated limitations include coarse LLM-based plausibility scoring, purely descriptive analysis, only three domains, and a small recovery benchmark.

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

🎓Beijing University of Posts and TelecommunicationsBeijing · China78 works · 33,365 citations

 Fields, topics and concepts

📚 Shelves 5The 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 worksComputer Vision and Pattern Recognition · 3,486 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,894 works

Cite

Cite as Yu Yan, Chunhong Zhang, Haiyu Zhao, Ziyang Zeng, Zihao Liu, Yongkang Wu, Jianzhou Diao, YiJie Chen, Shujie Wang and Zheng Hu. 2026. Beyond Noise: Characterizing Creative Potential in Unverifiable LLM Hallucinations. 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.554

@inproceedings{yan2026beyond,
  title = {{Beyond Noise: Characterizing Creative Potential in Unverifiable LLM Hallucinations}},
  author = {Yu Yan, Chunhong Zhang, Haiyu Zhao, Ziyang Zeng, Zihao Liu, Yongkang Wu, Jianzhou Diao, YiJie Chen, Shujie Wang and Zheng Hu},
  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.554},
  url = {https://doi.org/10.18653/v1/2026.acl-long.554}
}

References

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

  1. A. Aimi, G. Di Credico, H. Gimperlein, and C. Guarda-soni. 2025. Adaptive time-domain boundary element methods for the wave equation with neumann boundary conditions. Computers & Mathematics with Applications, 198:196–213. find
  2. Viraj Bagal, Rishal Aggarwal, PK Vinod, and U Deva Priyakumar. 2022. Molgpt: molecular generation using a transformer-decoder model. Journal of chemical information and modeling, 62(9):2064–2076. find
  3. Eranga Bandara, Safdar H. Bouk, Sachin Shetty, Sandip Roy, Ravi Mukkamala, Abdul Rahman, Peter Foytik, Xueping Liang, Ng Wee Keong, and Kasun De Zoysa. 2025. Llama-recipe — fine-tuned meta’s llama llm, pbom and nft enabled 5g network-slice orchestration and end-to-end supply-chain verification platf… find
  4. Yejin Bang, Samuel Cahyawijaya, Nayeon Lee, Wenliang Dai, Dan Su, Bryan Wilie, Holy Lovenia, Ziwei Ji, Tiezheng Yu, Willy Chung, Quyet V. Do, Yan Xu, and Pascale Fung. 2023. A multitask, multilingual, multimodal evaluation of ChatGPT on reasoning, hallucination, and interactivity. In Proceedings of…
  5. Iz Beltagy, Kyle Lo, and Arman Cohan. 2019. SciB-ERT: A pretrained language model for scientific text. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 3615– 3620,… find
  6. SUSAN P. BESEMER and DONALD J. TREFFINGER. 1981. Analysis of creative products: Review and synthesis. The Journal of Creative Behavior, 15(3):158– 178. find
  7. Thomas Bonnafont, Benjamin Chauvel, and Abdel-malek Toumi. 2024. A deep split-step wavelet model for the long-range propagation. In 2024 18th European Conference on Antennas and Propagation (EuCAP), pages 1–5. find
  8. Markus J. Buehler. 2024. Accelerating scientific discovery with generative knowledge extraction, graph-based representation, and multimodal intelligent graph reasoning. Preprint, arXiv:2403.11996. — [arXiv:2403.11996](https://arxiv.org/abs/2403.11996) find
  9. Tolga Buz, Benjamin Frost, Nikola Genchev, Moritz Schneider, Lucie-Aimée Kaffee, and Gerard de Melo. 2024. Investigating wit, creativity, and detectability of large language models in domain-specific writing style adaptation of Reddit’s showerthoughts. In Proceedings of the 13th Joint Conference on… find
  10. Nicholas Carlini, Florian Tramèr, Eric Wallace, Matthew Jagielski, Ariel Herbert-Voss, Katherine Lee, Adam Roberts, Tom Brown, Dawn Song, Úlfar Erlingsson, Alina Oprea, and Colin Raffel. 2021. Extracting training data from large language models. In 30th USENIX Security Symposium (USENIX Security 21)…
  11. Haoye Chai, Yuan Yuan, and Yong Li. 2025. Mobiworld: World models for mobile wireless network. Preprint, arXiv:2507.09462. — [arXiv:2507.09462](https://arxiv.org/abs/2507.09462) find
  12. Harrison Chase. 2022. LangChain. find
  13. Ioana Ciucă, Yuan-Sen Ting, Sandor Kruk, and Kartheik Iyer. 2023. Harnessing the power of adversarial prompting and large language models for robust hypothesis generation in astronomy. Preprint, arXiv:2306.11648. — [arXiv:2306.11648](https://arxiv.org/abs/2306.11648) find
  14. Elizabeth Clark, Tal August, Sofia Serrano, Nikita Haduong, Suchin Gururangan, and Noah A. Smith. 2021. All that’s ‘human’ is not gold: Evaluating human evaluation of generated text. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International… find
  15. Matthew DeLorenzo, Vasudev Gohil, and Jeyavijayan Rajendran. 2024. Creativeval: Evaluating creativity of llm-based hardware code generation. In 2024 IEEE LLM Aided Design Workshop (LAD), pages 1–5. IEEE. find
  16. Shehzaad Dhuliawala, Mojtaba Komeili, Jing Xu, Roberta Raileanu, Xian Li, Asli Celikyilmaz, and Jason Weston. 2024. Chain-of-verification reduces hallucination in large language models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 3563–3578, Bangkok, Thailand. Associ…
  17. Dayu Fan, Rui Meng, Song Gao, and Xiaodong Xu. 2025. Kgrag-sc: Knowledge graph rag-assisted semantic communication. Preprint, arXiv:2509.04801. — [arXiv:2509.04801](https://arxiv.org/abs/2509.04801) find
  18. Tao Feng, Yihang Sun, and Jiaxuan You. 2025. Grapheval: A lightweight graph-based llm framework for idea evaluation. Preprint, arXiv:2503.12600. find
  19. Tianyu Fu, Zihan Min, Hanling Zhang, Jichao Yan, Guohao Dai, Wanli Ouyang, and Yu Wang. 2026. Cache-to-cache: Direct semantic communication between large language models. Preprint, arXiv:2510.03215. — [arXiv:2510.03215](https://arxiv.org/abs/2510.03215) find
  20. Jiahui Geng, Fengyu Cai, Yuxia Wang, Heinz Koeppl, Preslav Nakov, and Iryna Gurevych. 2024. A survey of confidence estimation and calibration in large language models. In Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language…
  21. Carlos Gómez-Rodríguez and Paul Williams. 2023. A confederacy of models: a comprehensive evaluation of LLMs on creative writing. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 14504–14528, Singapore. Association for Computational Linguistics. find
  22. Maarten Grootendorst. 2020. Keybert: Minimal keyword extraction with bert. find
  23. Yunlong Gu, Meng Xu, Jiguang Li, Qilei Li, Zhao Huang, Mengshan Li, Lixin Guan, and Mikko Valkama. 2025. Wikan: Lightweight kolmogorov–arnold networks for accurate indoor wifi localization. Pervasive and Mobile Computing, 114:102121. find
  24. Joy Paul Guilford. 1950. Creativity. American psychologist, 5(9):444. find
  25. Sikun Guo, Amir Hassan Shariatmadari, Guangzhi Xiong, Albert Huang, Myles Kim, Corey M. Williams, Stefan Bekiranov, and Aidong Zhang. 2025. Ideabench: Benchmarking large language models for research idea generation. In Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Min… find
  26. Jiahui Hu, Dan Wang, Zhibo Wang, Xiaoyi Pang, Huiyu Xu, Ju Ren, and Kui Ren. 2025. Federated large language model: Solutions, challenges and future directions. IEEE Wireless Communications, 32(4):82–89. find
  27. 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 Transactions on Information Systems,…
  28. Ziwei Ji, Nayeon Lee, Rita Frieske, Tiezheng Yu, Dan Su, Yan Xu, Etsuko Ishii, Ye Jin Bang, Andrea Madotto, and Pascale Fung. 2023a. Survey of hallucination in natural language generation. ACM Computing Surveys, 55(12):1–38.
  29. Ziwei Ji, Tiezheng Yu, Yan Xu, Nayeon Lee, Etsuko Ishii, and Pascale Fung. 2023b. Towards mitigating LLM hallucination via self reflection. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 1827–1843, Singapore. Association for Computational Linguistics.
  30. Xuhui Jiang, Yuxing Tian, Fengrui Hua, Chengjin Xu, Yuanzhuo Wang, and Jian Guo. 2024. A survey on large language model hallucination via a creativity perspective. Preprint, arXiv:2402.06647. — [arXiv:2402.06647](https://arxiv.org/abs/2402.06647) find
  31. Faranak Khosravi and Mehdi Shadaram. 2025. Retentive time series: A scalable machine learning model for traffic prediction in elastic optical networks. IEEE Access, 13:116569–116585. find
  32. Yongjun Kim, Jihong Park, Mehdi Bennis, and Junil Choi. 2025. Resilient llm-empowered semantic mac protocols via zero-shot adaptation and knowledge distillation. Preprint, arXiv:2505.21518. — [arXiv:2505.21518](https://arxiv.org/abs/2505.21518) find
  33. Hirokazu Kiyomaru, Issa Sugiura, Daisuke Kawahara, and Sadao Kurohashi. 2024. A comprehensive analysis of memorization in large language models. In Proceedings of the 17th International Natural Language Generation Conference, pages 584–596, Tokyo, Japan. Association for Computational Linguistics. find
  34. Katherine Van Koevering and Jon Kleinberg. 2024. How random is random? evaluating the randomness and humaness of llms’ coin flips. Preprint, arXiv:2406.00092. — [arXiv:2406.00092](https://arxiv.org/abs/2406.00092) find
  35. Minhyeok Lee. 2023. A mathematical investigation of hallucination and creativity in gpt models. Mathematics, 11(10). find
  36. Junyi Li, Xiaoxue Cheng, Xin Zhao, Jian-Yun Nie, and Ji-Rong Wen. 2023. HaluEval: A large-scale hallucination evaluation benchmark for large language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 6449–6464, Singapore. Association for Computa…
  37. Guangming Liang, Mingjie Yang, Dongzhu Liu, Paul Henderson, and Lajos Hanzo. 2025. Environment-aware channel inference via cross-modal flow: From multimodal sensing to wireless channels. Preprint, arXiv:2512.04966. — [arXiv:2512.04966](https://arxiv.org/abs/2512.04966) find
  38. Peng Liao, Xuyu Wang, Yingxin Shan, Lingling An, and Shiwen Mao. 2025. Wireless sensing in artificial intelligence of things: A general quantum machine learning framework. IEEE Network, 39(3):207–214. find
  39. Xuanyu Liu, Shijian Gao, Boxun Liu, Xiang Cheng, and Liuqing Yang. 2025. Wifo-cf: Wireless foundation model for csi feedback. Preprint, arXiv:2508.04068. — [arXiv:2508.04068](https://arxiv.org/abs/2508.04068) find
  40. Yanzhen Liu, Zhijin Qin, Yongxu Zhu, and Geoffrey Ye Li. 2026. Enabling green wireless communications with neuromorphic continual learning. Preprint, arXiv:2502.17168. — [arXiv:2502.17168](https://arxiv.org/abs/2502.17168) find
  41. Potsawee Manakul, Adian Liusie, and Mark Gales. 2023. SelfCheckGPT: Zero-resource black-box hallucination detection for generative large language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 9004–9017, Singapore. Association for Computation… find
  42. Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Distributed representations of words and phrases and their compositionality. In Proceedings of the 27th International Conference on Neural Information Processing Systems - Volume 2, NIPS’13, page 3111–3119, Red Hook, NY,…
  43. OpenAI. 2024a. GPT-3.5 Turbo [large language model]. find
  44. OpenAI. 2024b. Gpt-4o system card. arXiv:2410.21276.
  45. Preprint, Kiho Park, Yo Joong Choe, and Victor Veitch. 2024. The linear representation hypothesis and the geometry of large language models. In Proceedings of the 41st International Conference on Machine Learning, ICML’24. JMLR.org.
  46. Biqing Qi, Kaiyan Zhang, Haoxiang Li, Kai Tian, Sihang Zeng, Zhang-Ren Chen, and Bowen Zhou. 2023. Large language models are zero shot hypothesis proposers. Preprint, arXiv:2311.05965. find
  47. Yansheng Qiu, Haoquan Zhang, Zhaopan Xu, Ming Li, Diping Song, Zheng Wang, and Kaipeng Zhang. 2025. Ai idea bench 2025: Ai research idea generation benchmark. Preprint, arXiv:2504.14191. — [arXiv:2504.14191](https://arxiv.org/abs/2504.14191) find
  48. Qwen Team. 2025. Qwen2.5 technical report. Preprint, arXiv:2412.15115. find
  49. Marissa Radensky, Simra Shahid, Raymond Fok, Pao Siangliulue, Tom Hope, and Daniel S. Weld. 2026. Human-llm compound system for scientific ideation through facet recombination and novelty evaluation. Preprint, arXiv:2409.14634. — [arXiv:2409.14634](https://arxiv.org/abs/2409.14634) find
  50. Kai Ruan, Xuan Wang, Jixiang Hong, Peng Wang, Yang Liu, and Hao Sun. 2026. Evaluating llms’ divergent thinking capabilities for scientific idea generation with minimal context. Preprint, arXiv:2412.17596.
  51. Huaguang Shi, Kaibo Jin, Xiaoquan Ren, Wei Li, and Yi Zhou. 2026. Channelmamba: A mamba-driven selective state-space model for channel prediction of high-mobility mimo in 6g iot. IEEE Transactions on Wireless Communications, 25:5291–5305. find
  52. Jung-Eun Shin, Adam J Riesselman, Aaron W Kol-lasch, Conor McMahon, Elana Simon, Chris Sander, Aashish Manglik, Andrew C Kruse, and Debora S Marks. 2021. Protein design and variant prediction using autoregressive generative models. Nature communications, 12(1):2403. find
  53. Chenglei Si, Diyi Yang, and Tatsunori Hashimoto. 2024. Can llms generate novel research ideas? a large-scale human study with 100+ nlp researchers. Preprint, arXiv:2409.04109.
  54. Zhihao Tao and Athina P. Petropulu. 2025. Meta-learning-driven gflownets for 3d directional modulation in mobile wireless systems. Preprint, arXiv:2511.06188. — [arXiv:2511.06188](https://arxiv.org/abs/2511.06188) find
  55. Jingwen Tong, Zijian Li, Fang Liu, Wei Guo, and Jun Zhang. 2026. Wirelessagent++: Automated agentic workflow design and benchmarking for wireless networks. Preprint, arXiv:2603.00501. — [arXiv:2603.00501](https://arxiv.org/abs/2603.00501) find
  56. Jingwen Tong, Jiawei Shao, Qiong Wu, Wei Guo, Zijian Li, Zehong Lin, and Jun Zhang. 2024. Wirelessagent: Large language model agents for intelligent wireless networks. Preprint, arXiv:2409.07964. — [arXiv:2409.07964](https://arxiv.org/abs/2409.07964) find
  57. Chris van der Lee, Albert Gatt, Emiel van Miltenburg, Sander Wubben, and Emiel Krahmer. 2019. Best practices for the human evaluation of automatically generated text. In Proceedings of the 12th International Conference on Natural Language Generation, pages 355–368, Tokyo, Japan. Association for Comp… find
  58. Haoyang Wang, Jiaxing He, and Lizhen Cui. 2025a. An advanced indoor localization method based on xlstm and residual multimodal fusion of uwb/imu data. Electronics, 14(13). find
  59. Xudong Wang, Jiacheng Wang, Lei Feng, Dusit Niyato, Ruichen Zhang, Jiawen Kang, Zehui Xiong, Hongyang Du, and Shiwen Mao. 2025b. Wireless hallucination in generative ai-enabled communications: Concepts, issues, and solutions. Preprint, arXiv:2503.06149. — [arXiv:2503.06149](https://arxiv.org/abs/2503.06149) find
  60. Xuesong Wang, Mo Li, Xingyan Shi, Zhaoqian Liu, and Shenghao Yang. 2025c. Diffusion-aided task-oriented semantic communications with model inversion attack. Preprint, arXiv:2506.19886. — [arXiv:2506.19886](https://arxiv.org/abs/2506.19886) find
  61. Zonglin Yang, Xinya Du, Junxian Li, Jie Zheng, Soujanya Poria, and Erik Cambria. 2024. Large language models for automated open-domain scientific hypotheses discovery. In Findings of the Association for Computational Linguistics: ACL 2024, pages 13545–13565, Bangkok, Thailand. Association for Comput…
  62. Yucheng Yu, Peng Chen, Xiao-Wei Zhu, Jianfeng Zhai, and Chao Yu. 2022. Continual learning digital predistortion of rf power amplifier for 6g ai-empowered wireless communication. IEEE Transactions on Microwave Theory and Techniques, 70(11):4916–4927. find
  63. Shuzhou Yuan, Zhan Qu, Ashish Yashwanth Kan-gen, and Michael Färber. 2025. Can hallucinations help? boosting llms for drug discovery. Preprint, arXiv:2501.13824. — [arXiv:2501.13824](https://arxiv.org/abs/2501.13824) find
  64. Xinran Zhao, Boyuan Zheng, Chenglei Si, Haofei Yu, Ken Liu, Runlong Zhou, Ruochen Li, Tong Chen, Xiang Li, Yiming Zhang, and Tongshuang Wu. 2025. The ramon llull’s thinking machine for automated ideation. Preprint, arXiv:2508.19200. — [arXiv:2508.19200](https://arxiv.org/abs/2508.19200) find
  65. Fenghao Zhu, Xinquan Wang, Chongwen Huang, Richeng Jin, Qianqian Yang, Ahmed Al Hammadi, Zhaoyang Zhang, Chau Yuen, and Mérouane Debbah. 2024. Robust continuous-time beam tracking with liquid neural network. In GLOBECOM 2024 - 2024 IEEE Global Communications Conference, page 4878–4883. IEEE. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yan-2026-beyond-noise-characterizing-creative-potential/