Uses WildChat
Show or Tell? Modeling the evolution of request-making in Human-LLM conversations
Findings of EACL 2026 · First arXiv submission
Introduces ReCCRE, which separates request content, context, assigned roles, and general expressions in user prompts. The study tracks how users phrase requests as they gain experience and over calendar time.
Data used
211,414 English request-making first turns from 18,964 inferred users, drawn from WildChat conversations collected between April 2023 and May 2024.
Findings
Users include more context in their requests over time, and individual users’ expression patterns tend to stabilize with experience.
Sections 3–5: segmentation, filtering, and longitudinal analyses. Read the source.
Dates and versions
First arXiv version. ArXiv version history.
Findings of EACL 2026 proceedings paper.
Cite this paper
Shengqi Zhu, Jeffrey M. Rzeszotarski, David Mimno. 2026. Show or Tell? Modeling the evolution of request-making in Human-LLM conversations. Findings of the Association for Computational Linguistics: EACL 2026.
Show BibTeX
@inproceedings{showortell2025,
title = {{Show or Tell? Modeling the evolution of request-making in Human-LLM conversations}},
author = {Zhu, Shengqi and Rzeszotarski, Jeffrey M. and Mimno, David},
year = {2026},
booktitle = {Findings of the Association for Computational Linguistics: EACL 2026},
publisher = {Association for Computational Linguistics},
pages = {5023--5034},
doi = {10.18653/v1/2026.findings-eacl.265},
url = {https://aclanthology.org/2026.findings-eacl.265/}
}