{"id":"dataset-signatures","title":"Dataset Signatures in Human-LLM Interactions and User Modeling","authors":[{"given":"Joseph","family":"Suh"},{"given":"Serina","family":"Chang"}],"first_public":"2026-10-04","date_note":"First arXiv version.","date_source":"https://arxiv.org/abs/2610.05534","year":2026,"venue":"arXiv preprint","paper_url":"https://arxiv.org/abs/2610.05534","category":"evaluation","relationship":"uses","dataset_version":"Seven conversation datasets, including WildChat-1M and WildChat-4.8M, with controlled comparisons to LMSYS-Chat-1M, ShareGPT, and other interaction sources.","summary":"Studies how conversation sources differ and how those differences propagate into trained user simulators and assistant evaluations. Uses the AI Observatory taxonomy to compare dataset categories alongside learned source classifiers.","finding":"Classifiers identify the source from user messages above chance even after taxonomy matching. Dataset choice changes user-model behavior and downstream evaluations.","evidence_url":"https://arxiv.org/html/2610.05534v1","evidence_note":"Sections 2–3 and Appendices A–C: dataset sampling, source classifiers, taxonomy matching, and user modeling. Cites WildChat, WildBench, and AI Observatory.","bibtex_key":"datasetsignatures2026","record_url":"https://yuntiandeng.com/wildchat/papers/dataset-signatures/","relationship_label":"Uses WildChat","citation_title":"Dataset Signatures in Human-LLM Interactions and User Modeling","citation_type":"article-journal","bibtex_type":"misc","citation_url":"https://arxiv.org/abs/2610.05534","citation_year":2026,"citation_note":"arXiv preprint.","arxiv":"2610.05534","doi":"10.48550/arXiv.2610.05534"}
