aiqbench analysis
This arXiv paper studies "model curiosity," the tendency of LLMs to ask epistemically useful questions that expose knowledge gaps and assumptions. It finds such questions are rare (17-28% of model questions across four datasets) but tend to prompt more exploratory user responses, with the benefit varying by context, such as advice, creative, and research tasks. For readers, the main takeaway is that asking better questions is a measurable behavior that could be trained for, though the evidence comes from a small set of datasets and a human study.
What Makes a Question Curious? Measuring Model Curiosity in Human-LLM Interactions
arXiv:2610.33950v1 Announce Type: cross Abstract: Large language models increasingly act as partners in everyday communication. In these interactions, useful assistance may require not only answering, but also asking the right questions that surface knowledge gaps, assumptions, and alternatives. In this study, we refer to such epistemic question-asking as model curiosity, and introduce a theory-grounded taxonomy that distinguishes curiosity-oriented questions from generic model questioning. Across four conversation datasets, we find that curiosity-oriented questions remain rare (17-28% of all questions), but are associated with more exploratory and epistemically rich subsequent user responses, with varying patterns across datasets. We further conduct a human study to evaluate the usefulness of model curiosity, finding that curiosity can encourage exploration, especially in advice, creative, and research contexts. These findings have implications for building LLMs that ask more selectively to support deeper exploration and human-AI collaboration.
Title:What Makes a Question Curious? Measuring Model Curiosity in Human-LLM Interactions
Abstract:Large language models increasingly act as partners in everyday communication. In these interactions, useful assistance may require not only answering, but also asking the right questions that surface knowledge gaps, assumptions, and alternatives. In this study, we refer to such epistemic question-asking as model curiosity, and introduce a theory-grounded taxonomy that distinguishes curiosity-oriented questions from generic model questioning. Across four conversation datasets, we find that curiosity-oriented questions remain rare (17-28% of all questions), but are associated with more exploratory and epistemically rich subsequent user responses, with varying patterns across datasets. We further conduct a human study to evaluate the usefulness of model curiosity, finding that curiosity can encourage exploration, especially in advice, creative, and research contexts. These findings have implications for building LLMs that ask more selectively to support deeper exploration and human-AI collaboration.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2610.33950 [cs.CL]
(or arXiv:2610.33950v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2610.33950 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 243 words
Related coverage
Qubit error correction blocked by common measurement shortcut
Quantum Zeitgeist · Oct 1, 2026
Interactions Between Crosscoder Features: A Compact Proofs Perspective
arXiv · Jun 10, 2026
Choose Your Game Wisely: Measuring Game-Theoretic Structures in Real-World Vehicle Interactions
arXiv · Aug 27, 2026
IDRBench: Benchmarking the Interactive Capabilities of Deep Research Agents
arXiv · Sep 25, 2026