Generative Interfaces transform queries into structured and interactive interfaces, offering richer organization and engagement than static conversational responses.
Large language models (LLMs) are increasingly seen as assistants, copilots, and consultants, capable of supporting a wide range of tasks through natural conversation. However, most systems remain constrained by a linear request-response format that often makes interactions inefficient in multi-turn, information-dense, and exploratory tasks. To address these limitations, we propose Generative Interfaces for Language Models, a paradigm in which LLMs respond to user queries by proactively generating user interfaces (UIs) that enable more adaptive and interactive engagement. Our framework leverages structured interface-specific representations and iterative refinements to translate user queries into task-specific UIs. For systematic evaluation, we introduce a multidimensional assessment framework that compares generative interfaces with traditional chat-based ones across diverse tasks, interaction patterns, and query types, capturing functional, interactive, and emotional aspects of user experience. Results show that generative interfaces consistently outperform conversational ones, with humans preferring them in over 70% of cases. These findings clarify when and why users favor generative interfaces, paving the way for future advancements in human-AI interaction.
User queries are first mapped into intermediate representations and then decoded into UI code. An iterative refinement process applies adaptive reward functions to optimize the generated interfaces.
Generative Interfaces consistently outperform conversational interfaces in human preference. Human feedback reveals key drivers of such preferences, such as cognitive offloading in complex scenarios and enhanced perceived credibility through structured presentation.
Human preferences vary across different domains, suggesting a balance between generative interfaces and conversational ones to optimize user experience.
@misc{chen2025generative,
title={Generative Interfaces for Language Models},
author={Jiaqi Chen and Yanzhe Zhang and Yutong Zhang and Yijia Shao and Diyi Yang},
year={2025},
eprint={2508.19227},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
Human Comments
"GenUI has a better webpage design, allowing the user to easily look for the information they want quickly."
"The design of the UI looks so much better in GenUI, it looks much cleaner and well balanced. The template selection screen is much more attractive and usable as well."
"GenUI has a better UI and user experience than the other one. The interactive elements make it much easier to navigate and understand."
"GenUI provides a much more detailed and clear example, making it easier to accomplish the task at hand."
"The layout and format looks very clean and nice on the eyes. It's intuitive and helps me focus on what's important."
"GenUI has a much better structure that practically holds the user's hand throughout the process, making it step-by-step."
"GenUI has a better webpage design, allowing the user to easily look for the information they want quickly."
"The design of the UI looks so much better in GenUI, it looks much cleaner and well balanced. The template selection screen is much more attractive and usable as well."
"GenUI has a better UI and user experience than the other one. The interactive elements make it much easier to navigate and understand."
"GenUI provides a much more detailed and clear example, making it easier to accomplish the task at hand."
"The layout and format looks very clean and nice on the eyes. It's intuitive and helps me focus on what's important."
"GenUI has a much better structure that practically holds the user's hand throughout the process, making it step-by-step."