Elamur Research⌕
☰
home / papers / 2026 / LAFaCT: Attribution-based Localization and Focused Sequential Analysis of Fact-Critical Tokens for Hallucination Detection

LAFaCT: Attribution-based Localization and Focused Sequential Analysis of Fact-Critical Tokens for Hallucination Detection

Paper · 2026 · Xin Wang, Jiahao Li, Licheng Zhang et al.

Summary

The paper tackles hallucination detection in LLMs using white-box access to hidden states. The authors argue that existing white-box detectors are poorly targeted. Some probe a single fixed token, such as the last one, and miss signals spread across the sequence. Others pool all tokens, so irrelevant tokens add noise.

LAFaCT addresses this in two steps, localizing first and analyzing second.

  • Localize: A lightweight proxy classifier is trained on a middle-layer hidden state of the final token. Its factuality probability is then traced back to the input embeddings with DeepLIFT. This yields per-token Factual Criticality scores, and the highest-scoring tokens are kept through Top-p selection.
  • Analyze: A sequential encoder processes only the hidden states of these selected tokens. It is trained with a new Angular Triplet Loss.

Across eight benchmarks covering QA, summarization and biography generation, and on several model families, LAFaCT outperforms prior detectors. It is especially strong on reasoning-heavy data and out-of-domain settings. The work matters because it shows that attending selectively to fact-bearing tokens gives more reliable factuality signals than probing one token or aggregating everything.

Key points

  • Architecture: The proxy is a two-layer MLP trained with Binary Cross-Entropy. DeepLIFT uses <pad> embeddings as its baseline. Token scores are L2-norms of the attributions, softmax-normalized over the response. The Analyze stage is a Bi-GRU with relative positional encoding.
  • Main result: On Llama-2-chat-hf-7B, LAFaCT beats the strongest baseline by an average of 3.2 AUROC points.
  • Ablations (Table 7): The full model scores 88.23 and 88.50 on TruthfulQA, and 87.07 and 89.25 on GSM8K, for Llama2 and Llama3 respectively. Angular Triplet Loss beats both standard Triplet Loss and CE loss, for example by 0.87 on GSM8K with Llama3.
  • Validating the token scores: On TruthfulQA, tokens that GPT-4 annotated as fact-critical receive clearly higher Factual Criticality scores. A standalone proxy classifier performs well below the full framework, regardless of proxy architecture.
  • Limitations: The method is supervised and needs labeled data. Labels for open-ended generation come from heuristics such as AlignScore thresholding, which may introduce noise.

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

🎓University of Science and Technology of ChinaHefei · China108 works · 112,698 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

Cite

Cite as Xin Wang, Jiahao Li, Licheng Zhang and Zhendong Mao. 2026. LAFaCT: Attribution-based Localization and Focused Sequential Analysis of Fact-Critical Tokens for Hallucination Detection. 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.312

@inproceedings{wang2026lafact,
  title = {{LAFaCT: Attribution-based Localization and Focused Sequential Analysis of Fact-Critical Tokens for Hallucination Detection}},
  author = {Xin Wang, Jiahao Li, Licheng Zhang and Zhendong Mao},
  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.312},
  url = {https://doi.org/10.18653/v1/2026.acl-long.312}
}

References

59 references, 29 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774.
  2. Amos Azaria and Tom Mitchell. 2023. The internal state of an llm knows when it’s lying. In The 2023 Conference on Empirical Methods in Natural Language Processing.
  3. Asma Ben Abacha and Dina Demner-Fushman. 2019. A question-entailment approach to question answering. BMC Bioinform., 20(1):511:1–511:23. find
  4. Jakub Binkowski, Dawid Janiak, Albert Sawczyn, and 1 others. 2025. Hallucination detection in llms using spectral features of attention maps. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 24365–24396.
  5. Collin Burns, Haotian Ye, Dan Klein, and Jacob Steinhardt. 2022. Discovering latent knowledge in language models without supervision. In The Eleventh International Conference on Learning Representations. find
  6. 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. arXiv preprint arXiv:2402.03744. find
  7. Yuyan Chen, Qiang Fu, Yichen Yuan, Zhihao Wen, Ge Fan, Dayiheng Liu, Dongmei Zhang, Zhixu Li, and Yanghua Xiao. 2023. Hallucination detection: Robustly discerning reliable answers in large language models. In Proceedings of the 32nd ACM International Conference on Information and Knowledge Managemen… find
  8. Yi Cheng, Xiao Liang, Yeyun Gong, Wen Xiao, Song Wang, Yuji Zhang, Wenjun Hou, Kaishuai Xu, Wenge Liu, Wenjie Li, and 1 others. 2024. Integrative decoding: Improve factuality via implicit self-consistency. arXiv preprint arXiv:2410.01556. — [arXiv:2410.01556](https://arxiv.org/abs/2410.01556) find
  9. Junyoung Chung, Caglar Gulcehre, KyungHyun Cho, and Yoshua Bengio. 2014. Empirical evaluation of gated recurrent neural networks on sequence modeling. arXiv preprint arXiv:1412.3555. find
  10. 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. arXiv preprint arXiv:2110.14168.
  11. Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou. 2019. Arcface: Additive angular margin loss for deep face recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4690–4699. find
  12. Xuefeng Du, Chaowei Xiao, and Sharon Li. 2024. Haloscope: Harnessing unlabeled llm generations for hallucination detection. Advances in Neural Information Processing Systems, 37:102948–102972. find
  13. Jinhao Duan, Hao Cheng, Shiqi Wang, Alex Zavalny, Chenan Wang, Renjing Xu, Bhavya Kailkhura, and Kaidi Xu. 2024. Shifting attention to relevance: Towards the predictive uncertainty quantification of free-form large language models. In Proceedings of the 62nd Annual Meeting of the Association for Com… find
  14. 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 e-prints, pages arXiv–2407.
  15. Ekaterina Fadeeva, Aleksandr Rubashevskii, Artem Shelmanov, Sergey Petrakov, Haonan Li, Hamdy Mubarak, Evgenii Tsymbalov, Gleb Kuzmin, Alexander Panchenko, Timothy Baldwin, and 1 others. 2024. Fact-checking the output of large language models via token-level uncertainty quantification. In Findings o…
  16. Ekaterina Fadeeva, Roman Vashurin, Akim Tsvigun, Artem Vazhentsev, Sergey Petrakov, Kirill Fedyanin, Daniil Vasilev, Elizaveta Goncharova, Alexander Panchenko, Maxim Panov, and 1 others. 2023. Lm-polygraph: Uncertainty estimation for language models. In Proceedings of the 2023 Conference on Empirica… find
  17. Marina Fomicheva, Shuo Sun, Lisa Yankovskaya, Frédéric Blain, Francisco Guzmán, Mark Fishel, Nikolaos Aletras, Vishrav Chaudhary, and Lucia Specia. 2020. Unsupervised quality estimation for neural machine translation. Transactions of the Association for Computational Linguistics, 8:539–555.
  18. 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
  19. Nuno M Guerreiro, Elena Voita, and André FT Martins. 2023. Looking for a needle in a haystack: A comprehensive study of hallucinations in neural machine translation. In Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 1059–1075.
  20. Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. TruthfulQA: Measuring how models mimic human falsehoods. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3214–3252, Dublin, Ireland. Association for Computational Linguistics…
  21. Jinwen He, Yujia Gong, Zijin Lin, Cheng’an Wei, Yue Zhao, and Kai Chen. 2024. Llm factoscope: Uncovering llms’ factual discernment through measuring inner states. In Findings of the Association for Computational Linguistics ACL 2024, pages 10218– 10230. find
  22. Zhen Lin, Shubhendu Trivedi, and Jimeng Sun. 2024. Generating with confidence: Uncertainty quantification for black-box large language models. Transactions on Machine Learning Research. find
  23. Bairu Hou, Yang Zhang, Jacob Andreas, and Shiyu Chang. 2024. A probabilistic framework for llm hallucination detection via belief tree propagation. arXiv preprint arXiv:2406.06950.
  24. 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
  25. 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…
  26. Minchul Kim, Anil K. Jain, and Xiaoming Liu. 2022. Adaface: Quality adaptive margin for face recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 18750–18759. find
  27. Lorenz Kuhn, Yarin Gal, and Sebastian Farquhar. 2023. Semantic uncertainty: Linguistic invariances for uncertainty estimation in natural language generation. arXiv preprint arXiv:2302.09664. find
  28. Chuang Li, Bingnan Xing, Dongdong Huo, Qihui Zhou, Zhen Xu, and Yu Wang. 2025. Mixhd: A method for detecting hallucinations based on the internal state and output probability of large language models. In ICASSP 2025 - 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICA… find
  29. Junyi Li, Jie Chen, Ruiyang Ren, Xiaoxue Cheng, Xin Zhao, Jian-Yun Nie, and Ji-Rong Wen. 2024. The dawn after the dark: An empirical study on factuality hallucination in large language models. In ACL (1).
  30. 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.
  31. Andrey Malinin and Mark Gales. 2020. Uncertainty estimation in autoregressive structured prediction. In International Conference on Learning Representations. find
  32. Potsawee Manakul, Adian Liusie, and Mark Gales. 2023. Selfcheckgpt: Zero-resource black-box hallucination detection for generative large language models. In The 2023 Conference on Empirical Methods in Natural Language Processing. find
  33. 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. arXiv preprint arXiv:2305.14251.
  34. Abhika Mishra, Akari Asai, Vidhisha Balachandran, Yizhong Wang, Graham Neubig, Yulia Tsvetkov, and Hannaneh Hajishirzi. 2024. Fine-grained hallucination detection and editing for language models. In First Conference on Language Modeling.
  35. Niels Mündler, Jingxuan He, Slobodan Jenko, and Martin Vechev. 2023. Self-contradictory hallucinations of large language models: Evaluation, detection and mitigation. In The Twelfth International Conference on Learning Representations.
  36. Shashi Narayan, Shay Cohen, and Mirella Lapata. 2018. Don’t give me the details, just the summary! topic-aware convolutional neural networks for extreme summarization. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing.
  37. 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. find
  38. 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…
  39. Seongheon Park, Xuefeng Du, Min-Hsuan Yeh, Haobo Wang, and Yixuan Li. 2025. Steer llm latents for hallucination detection. arXiv preprint arXiv:2503.01917. — [arXiv:2503.01917](https://arxiv.org/abs/2503.01917) find
  40. Siva Reddy, Danqi Chen, and Christopher D. Manning. 2019. CoQA: A conversational question answering challenge. Transactions of the Association for Computational Linguistics, 7:249–266.
  41. Florian Schroff, Dmitry Kalenichenko, and James Philbin. 2015. Facenet: A unified embedding for face recognition and clustering. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 815–823.
  42. Ramprasaath R. Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. 2017. Grad-CAM: Visual explanations from deep networks via gradient-based localization. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), pages 618– 626.
  43. Avanti Shrikumar, Peyton Greenside, and Anshul Kundaje. 2017. Learning important features through propagating activation differences. In International conference on machine learning, pages 3145–3153. PMlR.
  44. Ben Snyder, Marius Moisescu, and Muhammad Bilal Zafar. 2024. On early detection of hallucinations in factual question answering. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 2721–2732. find
  45. Weihang Su, Changyue Wang, Qingyao Ai, Yiran Hu, Zhijing Wu, Yujia Zhou, and Yiqun Liu. 2024. Unsupervised real-time hallucination detection based on the internal states of large language models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 14379–14391, Bangkok, Thai…
  46. Mukund Sundararajan, Ankur Taly, and Qiqi Yan. 2017. Axiomatic attribution for deep networks. In International conference on machine learning, pages 3319– 3328. PMLR. find
  47. Qwen Team. 2024. Qwen2.5: A party of foundation models. find
  48. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, Wenyin Fu, and 49 others… — Llama 2: Open Foundation and Fine-Tuned Chat Models
  49. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems, 30.
  50. Artem Vazhentsev, Lyudmila Rvanova, Ivan Lazichny, Alexander Panchenko, Maxim Panov, Timothy Baldwin, and Artem Shelmanov. 2025. Token-level density-based uncertainty quantification methods for eliciting truthfulness of large language models. In Proceedings of the 2025 Conference of the Nations of t…
  51. Tianlong Wang, Xianfeng Jiao, Yinghao Zhu, Zhongzhi Chen, Yifan He, Xu Chu, Junyi Gao, Yasha Wang, and Liantao Ma. 2025. Adaptive activation steering: A tuning-free llm truthfulness improvement method for diverse hallucinations categories. In Proceedings of the ACM on Web Conference 2025, WWW ’25, p… find
  52. Min-Hsuan Yeh, Max Kamachee, Seongheon Park, and Yixuan Li. 2024. Can your uncertainty scores detect hallucinated entity? In ICLR Workshop: Quantify Uncertainty and Hallucination in Foundation Models: The Next Frontier in Reliable AI. find
  53. Yakir Yehuda, Itzik Malkiel, Oren Barkan, Jonathan Weill, Royi Ronen, and Noam Koenigstein. 2024. Interrogatellm: Zero-resource hallucination detection in llm-generated answers. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages… find
  54. Fan Yin, Jayanth Srinivasa, and Kai-Wei Chang. 2024. Characterizing truthfulness in large language model generations with local intrinsic dimension. arXiv preprint arXiv:2402.18048. — [arXiv:2402.18048](https://arxiv.org/abs/2402.18048) find
  55. Mert Yüksekgönül, Varun Chandrasekaran, Erik Jones, Suriya Gunasekar, Ranjita Naik, Hamid Palangi, Ece Kamar, and Besmira Nushi. 2024. Attention satisfies: A constraint-satisfaction lens on factual errors of language models. In ICLR. find
  56. Yuheng Zha, Yichi Yang, Ruichen Li, and Zhiting Hu. 2023. AlignScore: Evaluating factual consistency with a unified alignment function. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 11328–11348, Toronto, Canada. Association…
  57. Tianhang Zhang, Lin Qiu, Qipeng Guo, Cheng Deng, Yue Zhang, Zheng Zhang, Chenghu Zhou, Xinbing Wang, and Luoyi Fu. 2023. Enhancing uncertainty-based hallucination detection with stronger focus. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 915– 932. find
  58. 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 1–45.
  59. Derui Zhu, Dingfan Chen, Qing Li, Zongxiong Chen, Lei Ma, Jens Grossklags, and Mario Fritz. 2024. Pollmgraph: Unraveling hallucinations in large language models via state transition dynamics. In Findings of the Association for Computational Linguistics: NAACL 2024, pages 4737–4751. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-lafact-attribution-based-localization-focused/