# Beyond ChatBots: ExploreLLM for Structured Thoughts and Personalized Model Responses

**Authors:** Xiao Ma, Swaroop Mishra, Ariel Liu, Sophie Su, Jilin Chen, Chinmay Kulkarni, Heng-Tze Cheng, Quoc Le, Ed Chi  
**arXiv:** [2312.00763](https://arxiv.org/abs/2312.00763)  
**Date:** 2023-12-01  
**Summary page:** https://www.summarizepaper.com/en/paper/2312.00763v1/

## Key points

- The paper introduces ExploreLLM, a system to enhance large language model (LLM) powered chatbots.
- Text-based chatbots lack structure and informational cues for exploratory or sensemaking tasks.
- ExploreLLM provides tools for structuring thoughts, exploring options, and navigating choices.
- Users can personalize responses by specifying high-level preferences.
- A user study showed positive results, indicating that ExploreLLM is helpful for planning tasks and personalizing responses.
- LLMs can be utilized beyond traditional chatbot interfaces by integrating natural language with graphical user interfaces.

## Summary

The paper titled "Beyond ChatBots: ExploreLLM for Structured Thoughts and Personalized Model Responses" introduces ExploreLLM, a system that aims to enhance the capabilities of large language model (LLM) powered chatbots. Currently, text-based chatbots impose a significant cognitive load on users when it comes to exploratory or sensemaking tasks like planning a trip or learning about a new city due to the lack of structure, informational cues, and the ability to specify high-level preferences or goals. ExploreLLM addresses these limitations by providing users with tools to structure their thoughts, explore different options, navigate through choices and recommendations, and steer models towards generating more personalized responses. The system incorporates a schema-like structure into tasks which guides users in planning and making the process more efficient. Additionally, ExploreLLM enables users to easily personalize responses by specifying high-level preferences. To evaluate the effectiveness of ExploreLLM, a user study was conducted with positive results indicating that users found it helpful for exploratory or planning tasks as it provides scaffolding and guidance throughout the process. Furthermore, participants were able to personalize responses more easily using ExploreLLM. Overall, this research suggests that LLMs can be utilized beyond traditional chatbot interfaces. By integrating natural language with graphical user interfaces and supporting complex user tasks, systems like ExploreLLM have the potential to revolutionize how users interact with LLM-powered models in various domains.