--- title: "Understanding and Detecting Hallucinations in Neural Machine" description: "Neural machine translation systems sometimes produce \"detached hallucinations,\" which are outputs… 1 reference, cited by 32 papers here." authors: ["Weijia Xu", "Sweta Agrawal", "Eleftheria Briakou", "Marianna J. Martindale", "Marine Carpuat"] year: 2023 venue: "Transactions of the Association for Computational Linguistics" source: "https://aclanthology.org/2023.tacl-1.32/" doi: "10.1162/tacl_a_00563" licence: "CC BY 4.0 (ACL Anthology)" licence_url: "https://creativecommons.org/licenses/by/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/xu-2023-understanding-detecting-hallucinations-neural-machine/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/xu-2023-understanding-detecting-hallucinations-neural-machine/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2023](https://md.research.elamur.ai/free/en/papers/2023/page/9/) / Understanding and Detecting Hallucinations in Neural Machine Translation via Model Introspection # Understanding and Detecting Hallucinations in Neural Machine Translation via Model Introspection **Authors:** [Weijia Xu](https://md.research.elamur.ai/free/en/authors/weijia-xu/), [Sweta Agrawal](https://md.research.elamur.ai/free/en/authors/sweta-agrawal/), [Eleftheria Briakou](https://md.research.elamur.ai/free/en/authors/eleftheria-briakou/), Marianna J. Martindale and [Marine Carpuat](https://md.research.elamur.ai/free/en/authors/marine-carpuat/) · **Year:** 2023 · **Venue:** [Transactions of the Association for Computational Linguistics](https://md.research.elamur.ai/free/en/venues/tacl/) ([Transactions of the Association for Computational Linguistics (TACL) 2023](https://md.research.elamur.ai/free/en/venues/tacl/2023/)) · **Publisher:** [The MIT Press](https://md.research.elamur.ai/free/en/publishers/the-mit-press/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2023.tacl-1.32/) · [PDF](https://aclanthology.org/2023.tacl-1.32.pdf) · [DOI 10.1162/tacl\_a\_00563](https://doi.org/10.1162/tacl_a_00563) · [OpenAlex](https://openalex.org/W4380995299) **Citations:** 53 citations (OpenAlex, 2026-10-14) · [32](https://md.research.elamur.ai/free/en/papers/xu-2023-understanding-detecting-hallucinations-neural-machine/cited-by/) in our library, of the 149,439 papers whose bibliographies we read ## Summary Neural machine translation systems sometimes produce "detached hallucinations," which are outputs largely unrelated to the source sentence. These errors are rare in natural text, so it has been hard to study when they occur, and most existing detectors treat the model as a black box. The authors look inside a Transformer NMT model using Layerwise Relevance Propagation (LRP), which measures how much each source and target token contributes to every prediction. They built an English→Chinese counterfactual dataset by perturbing source sentences and pairing hallucinated outputs with normal translations of the unperturbed inputs. On this data they tested three hypotheses about hallucination symptoms. Two source-side patterns proved to be the clearest signals: contributions concentrated on a few source tokens, and a contribution distribution that barely changes across decoding steps. The relative weight of source versus target context mattered less. Features built from these patterns feed a lightweight classifier. On manually annotated English-Chinese and German-English hallucinations, it beats model-free baselines, quality-estimation-based detectors, and black-box detectors built on large pre-trained models, and it holds up better under domain shift. ## Key points - **Perturbations used:** - character deletion (misspelling) with probability 0.1; - random title-casing with probability 0.1; - inserting one token at the start of the sentence, drawn from frequent, rare, or mid-frequency tokens, or punctuation. - **Data construction:** 50k seed sentence pairs were sampled from the training corpora. Hallucinations were flagged with BLEU-based heuristics: - the original translation must be close to the reference (BLEU > 0.3); - the perturbed-input output must be far from the original translation (BLEU \< 0.03); - the perturbed-input output must not simply copy the source (BLEU \< 0.5). - **Resulting dataset:** 623 contrastive pairs for misspelling, 270 for title-casing, and 1307 for insertion. - **Hypotheses tested:** Low Source Contribution (Voita et al.), Local Source Contribution (Lee et al.), and a newly formalized Static Source Contribution hypothesis. The two source-based patterns were stronger indicators than low source contribution. - **Resources and limitations:** The best detector, LRP + LASER, also looked promising in a small stress test where hallucinations were rare. The study covers only detached hallucinations and WMT-trained models. Code and data are available at github.com/weijia-xu/hallucinations-in-nmt. *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 - [Microsoft](https://md.research.elamur.ai/free/en/companies/microsoft/) — Company, Redmond, United States · 378 works · 161,630 citations - [University of Maryland, College Park](https://md.research.elamur.ai/free/en/universities/university-of-maryland-college-park/) — University, College Park, United States · 141 works · 29,174 citations 📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works 🧭 Topics (1): [🔁 Machine translation](https://md.research.elamur.ai/free/en/topics/machine-translation/) · 1,105 works ## Cite Cite as: Weijia Xu, Sweta Agrawal, Eleftheria Briakou, Marianna J. Martindale and Marine Carpuat. 2023. Understanding and Detecting Hallucinations in Neural Machine Translation via Model Introspection. Transactions of the Association for Computational Linguistics. The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @article{xu2023understanding, title = {{Understanding and Detecting Hallucinations in Neural Machine Translation via Model Introspection}}, author = {Weijia Xu, Sweta Agrawal, Eleftheria Briakou, Marianna J. Martindale and Marine Carpuat}, journal = {Transactions of the Association for Computational Linguistics}, year = {2023}, doi = {10.1162/tacl\_a\_00563}, url = {https://doi.org/10.1162/tacl_a_00563} } ``` - [← previous Investigating the Translation Performance of a Large Multilingual Language Model: the Case of BLOOM](https://md.research.elamur.ai/free/en/papers/bawden-2023-investigating-translation-performance-large-multilingual/) - [next → Bridging the Gap: A Survey on Integrating (Human) Feedback for Natural Language Generation](https://md.research.elamur.ai/free/en/papers/fernandes-2023-bridging-gap-survey-integrating-human/) - [↑ back to papers of 2023](https://md.research.elamur.ai/free/en/papers/2023/page/9/) ## References 1 references, 1 of them with a card of ours. ↗ leads to the work. 1. Farhad Akhbardeh, Arkady Arkhangorodsky, Magdalena Biesialska, Ondřej Bojar, Rajen Chatterjee, Vishrav Chaudhary, Marta R. Costa-jussa, Cristina España-Bonet, Angela Fan, Christian Federmann, Markus Freitag, Yvette Graham, Roman Grundkiewicz, Barry Haddow, Leonie Harter, Kenneth Heafield, Christop… — [Findings of the 2021 Conference on Machine Translation (WMT21)](https://md.research.elamur.ai/free/en/papers/akhbardeh-2021-findings-2021-conference-machine-translation/) *Updated 29 October 2026*