WildChat / research hub

Uses WildChat

QUILT: Query–User Interaction Linguistic Taxonomy

Yuyang Dai

AACL-IJCNLP 2026 · Conference begins

Introduces a joint taxonomy of task intent and query-level expertise expressed in language. Combines model annotation and human review, then evaluates classification and task–expertise-conditioned responses.

Data used

3,000 English-labeled WildChat dialogues, with task type and expressed expertise annotated from the first user turn.

Findings

Debugging queries are disproportionately labeled Expert, and Expert-labeled queries are longer within task categories. Joint task–expertise conditioning improves mean response-calibration scores. These labels describe linguistic evidence, not verified personal competence.

Sections 3–6 and appendices: annotation, analysis, classification, and response adaptation. The PDF defines query-level expertise and cites WildChat; this entry follows the full paper rather than the broader wording in the landing-page abstract. Read the source.

Dates and versions

The proceedings date the conference November 6–10, 2026; the Anthology lists publication in November 2026. Proceedings metadata.

AACL-IJCNLP 2026 proceedings paper.

Cite this paper

Yuyang Dai. 2026. QUILT: Query–User Interaction Linguistic Taxonomy. Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers).

Show BibTeX
@inproceedings{dai-2026-quilt,
  title = {{QUILT: Query–User Interaction Linguistic Taxonomy}},
  author = {Dai, Yuyang},
  year = {2026},
  booktitle = {Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)},
  publisher = {Association for Computational Linguistics},
  pages = {3018--3035},
  doi = {10.18653/2026.aacl-long.138},
  url = {https://aclanthology.org/2026.aacl-long.138/}
}

Reuse this entry