[{"_id":"66885","publisher":"ACL","language":[{"iso":"eng"}],"user_id":"76456","status":"public","title":"Syntactic complexity convergence in dialogue: Analysis of the phenomenon and application to LLM detection","year":"2026","author":[{"full_name":"Wang, Yu","last_name":"Wang","first_name":"Yu"},{"last_name":"Chen","first_name":"Yanran","full_name":"Chen, Yanran"},{"full_name":"Wang, Yifan","first_name":"Yifan","last_name":"Wang"},{"full_name":"Eger, Steffen","first_name":"Steffen","last_name":"Eger"},{"id":"76456","full_name":"Buschmeier, Hendrik","last_name":"Buschmeier","first_name":"Hendrik","orcid":"0000-0002-9613-5713"}],"conference":{"end_date":"2026-10-29","location":"Budapest, Hungary","name":"2026 Conference on Empirical Methods in Natural Language Processing","start_date":"2026-10-24"},"publication_status":"accepted","date_updated":"2026-08-31T17:33:32Z","date_created":"2026-08-31T17:27:49Z","place":"Budapest, Hungary","type":"conference","department":[{"_id":"660"}],"publication":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing","citation":{"chicago":"Wang, Yu, Yanran Chen, Yifan Wang, Steffen Eger, and Hendrik Buschmeier. “Syntactic Complexity Convergence in Dialogue: Analysis of the Phenomenon and Application to LLM Detection.” In <i>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</i>. Budapest, Hungary: ACL, n.d.","short":"Y. Wang, Y. Chen, Y. Wang, S. Eger, H. Buschmeier, in: Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, ACL, Budapest, Hungary, n.d.","ieee":"Y. Wang, Y. Chen, Y. Wang, S. Eger, and H. Buschmeier, “Syntactic complexity convergence in dialogue: Analysis of the phenomenon and application to LLM detection,” presented at the 2026 Conference on Empirical Methods in Natural Language Processing, Budapest, Hungary.","apa":"Wang, Y., Chen, Y., Wang, Y., Eger, S., &#38; Buschmeier, H. (n.d.). Syntactic complexity convergence in dialogue: Analysis of the phenomenon and application to LLM detection. <i>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</i>. 2026 Conference on Empirical Methods in Natural Language Processing, Budapest, Hungary.","bibtex":"@inproceedings{Wang_Chen_Wang_Eger_Buschmeier, place={Budapest, Hungary}, title={Syntactic complexity convergence in dialogue: Analysis of the phenomenon and application to LLM detection}, booktitle={Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing}, publisher={ACL}, author={Wang, Yu and Chen, Yanran and Wang, Yifan and Eger, Steffen and Buschmeier, Hendrik} }","ama":"Wang Y, Chen Y, Wang Y, Eger S, Buschmeier H. Syntactic complexity convergence in dialogue: Analysis of the phenomenon and application to LLM detection. In: <i>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</i>. ACL.","mla":"Wang, Yu, et al. “Syntactic Complexity Convergence in Dialogue: Analysis of the Phenomenon and Application to LLM Detection.” <i>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</i>, ACL."},"quality_controlled":"1","abstract":[{"text":"We revisit the under-explored phenomenon of syntactic complexity convergence in dialogue. Extending prior topic-level analyses to full conversations, we examine five human-human datasets and a set of human-LLM dialogues. We integrate and evaluate a multidimensional set of syntactic complexity metrics to quantify structural variation in dialogue. Our analyses reveal robust syntactic complexity convergence across human-human datasets, especially in task-oriented interactions, and statistical tests confirm that these effects are not due to chance. In contrast, human-LLM dialogues show no significant convergence under individual metrics; however, linear-CKA uncovers a distinct cross-metric convergence pattern. Motivated by these differences, we use syntactic complexity as a feature for detecting LLM-involved dialogue, achieving 96\\% accuracy. Our classification results demonstrate its potential without relying on likelihood-based signals such as surprisal, which require a language model as the estimator.","lang":"eng"}],"project":[{"_id":"2752","name":"TRR 318-2 - Project B07: Communicative practices of requesting information and explanation from LLM-based agents"},{"name":"TRR 318-2 - Project A02: Monitoring the understanding of explanations","_id":"2736"}]}]