Elamur Research⌕
☰
home / papers / 2026 / Two Pathways to Truthfulness: On the Intrinsic Encoding of LLM Hallucinations

Two Pathways to Truthfulness: On the Intrinsic Encoding of LLM Hallucinations

Paper · 2026 · Wen Luo, Guangyue Peng, Wei Li et al.

Summary

Probing classifiers trained on LLM hidden states can detect hallucinations, but it has been unclear where these truthfulness signals come from. Luo et al. propose that they arise through two separate routes. In the Question-Anchored pathway, the signal depends on information moving from the key question tokens to the answer. In the Answer-Anchored pathway, the generated answer carries its own evidence.

The hypothesis came from saliency analysis, which showed a bimodal pattern in how much detection relies on question-to-answer flow. The authors then tested it with attention knockout and token patching on 4 QA datasets and 12 models, spanning base, instruction-tuned and reasoning variants. They found that:

  • Question-Anchored encoding is typical for well-known facts, while Answer-Anchored encoding is typical for long-tail ones.
  • Hidden states themselves reveal which pathway is active.

Using these findings, they built two detectors, Mixture-of-Probes (MoP) and Pathway Reweighting (PR). Both beat existing internal-signal baselines, with gains of up to 10% AUC. The work links interpretability to more reliable self-monitoring of LLM outputs.

Key points

  • The datasets are PopQA, TriviaQA, HotpotQA and Natural Questions (NQ). The analysis focuses on "exact" question and answer tokens, which are defined using semantic frame theory.
  • Blocking question-to-answer attention badly hurts Question-Anchored probes, while Answer-Anchored probes stay largely robust.
  • MoP uses pathway-specialized expert probes, routed by a probe that recognizes which pathway is active. PR rescales attention between exact answer and question tokens through a lightweight adapter used only during detection.
  • In one reported configuration, PR reaches 94.01 AUC on PopQA, compared with 88.71 for the standard probing baseline. Ablations with random gating or vanilla experts perform worse than the full MoP.
  • The main limitation is that the methods require white-box access to model internals. Code is published at github.com/RowanWenLuo/llm-truthfulness-pathways.

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

🎓Peking UniversityBeijing · China244 works · 39,359 citations🏢Microsoft Research AsiaBeijing · China73 works · 60,539 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 1Research subjects and languages the work is about.Language models & pretraining · 3,320 works

Cite

Cite as Wen Luo, Guangyue Peng, Wei Li, Shaohang Wei, Feifan Song, Liang Wang, Nan Yang, Xingxing Zhang, Jing Jin, Furu Wei and Houfeng Wang. 2026. Two Pathways to Truthfulness: On the Intrinsic Encoding of 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.1173

@inproceedings{luo2026two,
  title = {{Two Pathways to Truthfulness: On the Intrinsic Encoding of LLM Hallucinations}},
  author = {Wen Luo, Guangyue Peng, Wei Li, Shaohang Wei, Feifan Song, Liang Wang, Nan Yang, Xingxing Zhang, Jing Jin, Furu Wei and Houfeng Wang},
  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.1173},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1173}
}

References

38 references, 16 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Lukas Aichberger, Kajetan Schweighofer, Mykyta Ielanskyi, and Sepp Hochreiter. 2024. Semantically diverse language generation for uncertainty estimation in language models. arXiv preprint arXiv:2406.04306. — [arXiv:2406.04306](https://arxiv.org/abs/2406.04306) find
  2. Lukas Aichberger, Kajetan Schweighofer, Mykyta Ielanskyi, and Sepp Hochreiter. 2025. Improving uncertainty estimation through semantically diverse language generation. In The Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 24-28, 2025. OpenReview.net. find
  3. Ge Bai, Jie Liu, Xingyuan Bu, Yancheng He, Jiaheng Liu, Zhanhui Zhou, Zhuoran Lin, Wenbo Su, Tiezheng Ge, Bo Zheng, and Wanli Ouyang. 2024. Mt-bench-101: A fine-grained benchmark for evaluating large language models in multi-turn dialogues. In Proceedings of the 62nd Annual Meeting of the Associatio…
  4. Collin F Baker, Charles J Fillmore, and John B Lowe. 1998. The berkeley framenet project. In 36th Annual Meeting of the Association for Computational Linguistics and 17th International Conference on Computational Linguistics, Volume 1, pages 86–90.
  5. Dylan Bouchard and Mohit Singh Chauhan. 2025. Uncertainty quantification for language models: A suite of black-box, white-box, llm judge, and ensemble scorers. arXiv preprint arXiv:2504.19254. — [arXiv:2504.19254](https://arxiv.org/abs/2504.19254) find
  6. Collin Burns, Haotian Ye, Dan Klein, and Jacob Steinhardt. 2023. Discovering latent knowledge in language models without supervision. In The Eleventh International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1-5, 2023. OpenReview.net. find
  7. Chao Chen, Kai Liu, Ze Chen, Yi Gu, Yue Wu, Mingyuan Tao, Zhihang Fu, and Jieping Ye. 2024. INSIDE: llms’ internal states retain the power of hallucination detection. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net. find
  8. Constanza Fierro, Negar Foroutan, Desmond Elliott, and Anders Søgaard. 2025. How do multilingual language models remember facts? In Findings of the Association for Computational Linguistics, ACL 2025, Vienna, Austria, July 27 - August 1, 2025, pages 16052–16106. Association for Computational Linguis… find
  9. Mor Geva, Jasmijn Bastings, Katja Filippova, and Amir Globerson. 2023. Dissecting recall of factual associations in auto-regressive language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023, pages 12216–1223… find
  10. Asma Ghandeharioun, Avi Caciularu, Adam Pearce, Lucas Dixon, and Mor Geva. 2024. Patchscopes: A unifying framework for inspecting hidden representations of language models. In Forty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024. OpenReview.net. find
  11. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, and 542 o… — The Llama 3 Herd of Models
  12. Wentao Hu, Wengyu Zhang, Yiyang Jiang, Chen Jason Zhang, Xiaoyong Wei, and Qing Li. 2025. Removal of hallucination on hallucination: Debate-augmented RAG. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2025, Vienna, Austria, Ju…
  13. Lei Huang, Xiaocheng Feng, Weitao Ma, Yuchun Fan, Xiachong Feng, Yuxuan Gu, Yangfan Ye, Liang Zhao, Weihong Zhong, Baoxin Wang, Dayong Wu, Guoping Hu, Lingpeng Kong, Tong Xiao, Ting Liu, and Bing Qin. 2025. Alleviating hallucinations from knowledge misalignment in large language models via selective… find
  14. Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix,… find
  15. Mandar Joshi, Eunsol Choi, Daniel Weld, and Luke Zettlemoyer. 2017. TriviaQA: A large scale distantly supervised challenge dataset for reading comprehension. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1601–1611, Vancouver…
  16. Saurav Kadavath, Tom Conerly, Amanda Askell, T. J. Henighan, Dawn Drain, Ethan Perez, Nicholas Schiefer, Zachary Dodds, Nova Dassarma, Eli Tran-Johnson, Scott Johnston, Sheer El-Showk, Andy Jones, Nelson Elhage, Tristan Hume, Anna Chen, Yuntao Bai, Sam Bowman, Stanislav Fort, and 17 others. 2022. La… find
  17. Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, Kristina Toutanova, Llion Jones, Matthew Kelcey, Ming-Wei Chang, Andrew M. Dai, Jakob Uszkoreit, Quoc Le, and Slav Petrov. 2019. Natural…
  18. Kenneth Li, Oam Patel, Fernanda B. Viégas, Hanspeter Pfister, and Martin Wattenberg. 2023. Inference-time intervention: Eliciting truthful answers from a language model. In Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS…
  19. Wen Luo, Tianshu Shen, Wei Li, Guangyue Peng, Richeng Xuan, Houfeng Wang, and Xi Yang. 2024. Halludial: A large-scale benchmark for automatic dialogue-level hallucination evaluation. Preprint, arXiv:2406.07070. — [arXiv:2406.07070](https://arxiv.org/abs/2406.07070) find
  20. Wen Luo, Feifan Song, Wei Li, Guangyue Peng, Shaohang Wei, and Houfeng Wang. 2025. Odysseus navigates the sirens’ song: Dynamic focus decoding for factual and diverse open-ended text generation. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Lon… find
  21. Alex Mallen, Akari Asai, Victor Zhong, Rajarshi Das, Daniel Khashabi, and Hannaneh Hajishirzi. 2023. When not to trust language models: Investigating effectiveness of parametric and non-parametric memories. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (V… find
  22. Paul Michel, Omer Levy, and Graham Neubig. 2019. Are sixteen heads really better than one? Advances in neural information processing systems, 32.
  23. Sewon Min, Kalpesh Krishna, Xinxi Lyu, Mike Lewis, Wen-tau Yih, Pang Wei Koh, Mohit Iyyer, Luke Zettlemoyer, and Hannaneh Hajishirzi. 2023. Factscore: Fine-grained atomic evaluation of factual precision in long form text generation. In Proceedings of the 2023 Conference on Empirical Methods in Natur…
  24. Mengjia Niu, Hamed Haddadi, and Guansong Pang. 2025. Robust hallucination detection in llms via adaptive token selection. arXiv preprint arXiv:2504.07863. — [arXiv:2504.07863](https://arxiv.org/abs/2504.07863) find
  25. Hadas Orgad, Michael Toker, Zorik Gekhman, Roi Reichart, Idan Szpektor, Hadas Kotek, and Yonatan Belinkov. 2025. Llms know more than they show: On the intrinsic representation of LLM hallucinations. In The Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 2… find
  26. Artidoro Pagnoni, Vidhisha Balachandran, and Yulia Tsvetkov. 2021. Understanding factuality in abstractive summarization with frank: A benchmark for factuality metrics. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Languag…
  27. Chen Qian, Dongrui Liu, Haochen Wen, Zhen Bai, Yong Liu, and Jing Shao. 2025. Demystifying reasoning dynamics with mutual information: Thinking tokens are information peaks in llm reasoning. arXiv preprint arXiv:2506.02867. — [arXiv:2506.02867](https://arxiv.org/abs/2506.02867) find
  28. Zhengliang Shi, Shuo Zhang, Weiwei Sun, Shen Gao, Pengjie Ren, Zhumin Chen, and Zhaochun Ren. 2024. Generate-then-ground in retrieval-augmented generation for multi-hop question answering. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Pape…
  29. Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. 2014. Deep inside convolutional networks: Visualising image classification models and saliency maps. In 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14-16, 2014, Workshop Track Proceedings.
  30. Yuanhe Tian, Ruyi Gan, Yan Song, Jiaxing Zhang, and Yongdong Zhang. 2024. Chimed-gpt: A chinese medical large language model with full training regime and better alignment to human preferences. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long… find
  31. Yue Zhang, Yafu Li, Leyang Cui, Deng Cai, Lemao Liu, Tingchen Fu, Xinting Huang, Enbo Zhao, Yu Zhang, Yulong Chen, and 1 others. 2025. Siren’s song in the ai ocean: A survey on hallucination in large language models. Computational Linguistics, pages 1–46.
  32. Eric Todd, Millicent L. Li, Arnab Sen Sharma, Aaron Mueller, Byron C. Wallace, and David Bau. 2024. Function vectors in large language models. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenReview.net. find
  33. Lean Wang, Lei Li, Damai Dai, Deli Chen, Hao Zhou, Fandong Meng, Jie Zhou, and Xu Sun. 2023. Label words are anchors: An information flow perspective for understanding in-context learning. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 9840–9855.
  34. Wenhao Wu, Yizhong Wang, Guangxuan Xiao, Hao Peng, and Yao Fu. 2025. Retrieval head mechanistically explains long-context factuality. In The Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 24-28, 2025. OpenReview.net.
  35. Boyang Xue, Fei Mi, Qi Zhu, Hongru Wang, Rui Wang, Sheng Wang, Erxin Yu, Xuming Hu, and Kam-Fai Wong. 2025a. UAlign: Leveraging uncertainty estimations for factuality alignment on large language models. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volum… find
  36. Yihao Xue, Kristjan Greenewald, Youssef Mroueh, and Baharan Mirzasoleiman. 2025b. Verify when uncertain: Beyond self-consistency in black box hallucination detection. arXiv preprint arXiv:2502.15845. — [arXiv:2502.15845](https://arxiv.org/abs/2502.15845) find
  37. 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 41 others. 2025. Qwen3 technical report. 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 find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/luo-2026-two-pathways-truthfulness-intrinsic-encoding/