AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Prompts

Wu, Tongshuang; Terry, Michael; Cai, Carrie J.

doi:10.1145/3491102.3517582

Cited by 155 publications

(78 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One of our future works is improving the quality of the generated query examples. Recently, generative language models [56] trained on a large corpus have been proposed to facilitate various downstream tasks [57], [58]. They open up the possibility of enhancing the diversity and naturalness of linguistic expression while maintaining the accurate representation of the users' intentions.…”

Section: Limitations and Future Workmentioning

confidence: 99%

XNLI: Explaining and Diagnosing NLI-Based Visual Data Analysis

Feng

Wang

Pan

et al. 2024

IEEE Trans. Visual. Comput. Graphics

View full text Add to dashboard Cite

Natural language interfaces (NLIs) enable users to flexibly specify analytical intentions in data visualization. However, diagnosing the visualization results without understanding the underlying generation process is challenging. Our research explores how to provide explanations for NLIs to help users locate the problems and further revise the queries. We present XNLI, an explainable NLI system for visual data analysis. The system introduces a Provenance Generator to reveal the detailed process of visual transformations, a suite of interactive widgets to support error adjustments, and a Hint Generator to provide query revision hints based on the analysis of user queries and interactions. Two usage scenarios of XNLI and a user study verify the effectiveness and usability of the system. Results suggest that XNLI can significantly enhance task accuracy without interrupting the NLI-based analysis process.

show abstract

Section: Limitations and Future Workmentioning

confidence: 99%

XNLI: Explaining and Diagnosing NLI-Based Visual Data Analysis

Feng

Wang

Pan

et al. 2024

IEEE Trans. Visual. Comput. Graphics

View full text Add to dashboard Cite

show abstract

“…Hence, if to use chatbots powered by GPT-3 to facilitate data collection, we suggest that important slots be put earlier in the prompt and the number of questions of specific data type be limited. If more data slots need to be collected, multi-stage prompts [104] can be considered.…”

Section: Designing Effective Prompts For Chatbots That Collect Self-r...mentioning

confidence: 99%

“…Measuring chatbots requires great human efforts so more future research into the effects of these parameters on prompts is needed to provide guidance for the development of better and more robust chatbots. Researchers can also investigate multi-stage prompting [99,104] (i.e., designing several prompts for different questions in one dialogue session) if they intend to collect more than 5 slots of information. Such approaches will require incorporating dialog state tracking techniques (e.g., [58]) for automated slot filling.…”

Section: Future Workmentioning

confidence: 99%

“…Despite these potentials, it is yet fully understood how LLMs read the prompt and use pretrained knowledge [14,59], the development of prompts is usually conducted through iterative trial and error [61]. While the HCI community have actively explored the use of LLMs in various domains (e.g., [20,49,104]), research that leverages LLMs for powering chatbots, particularly task-oriented ones [8,67,94], is still sparse. Due to the inherent characteristics of LLMs, LLM-driven chatbots may be error-prone [44] or digress from their tasks [94].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Leveraging Large Language Models to Power Chatbots for Collecting User Self-Reported Data

Wei¹,

Kim²,

Jung³

et al. 2023

Preprint

View full text Add to dashboard Cite

Large Language Model How's your day so far? Prefix 📄 Information format 🤖 Persona modifier A Ongoing Dialogue History Insert as prompt Generate chatbot response B Work & Productivity ✅ Work done ✅ Rate productivity ✅ Other concerns at work ✅ What went well a C Going splendid so far. How is yours? I'm doing well, thanks for asking. Let's talk about your work and productivity yesterday. What work did you get done?Figure 1: An overview of our chatbot running on a large language model through zero-shot response generation, with only a prefix A○ consisting of persona modifier and information format and the ongoing dialogue history B ○. The example conversation is carried on about work a ○. GPT-3 was used for an underlying large language model C ○. Refer to the online demo at https://naver-ai.github.io/llm-chatbot.

show abstract

“…Moreover, the resulting subquestions in least-to-most prompting are usually dependent on each other and have to be sequentially solved in a specific order so that answers to some subquestions can be used to as building blocks to solve other subquestions. Wu et al (2022) propose chaining large language model steps such that the output of one step becomes the input for the next and develop an interactive system for users to construct and modify chains. Least-to-most prompting chains the processes of problem reduction and subproblem solving.…”

Section: Related Workmentioning

confidence: 99%

Least-to-Most Prompting Enables Complex Reasoning in Large Language Models

Zhou¹,

Schärli²,

Hou³

et al. 2022

Preprint

View full text Add to dashboard Cite

We propose a novel prompting strategy, least-to-most prompting, that enables large language models to better perform multi-step reasoning tasks. Least-to-most prompting first reduces a complex problem into a list of subproblems, and then sequentially solves the subproblems, whereby solving a given subproblem is facilitated by the model's answers to previously solved subproblems. Experiments on symbolic manipulation, compositional generalization and numerical reasoning demonstrate that least-to-most prompting can generalize to examples that are harder than those seen in the prompt context, outperforming other promptingbased approaches by a large margin. A notable empirical result is that the GPT-3 code-davinci-002 model with least-to-most-prompting can solve the SCAN benchmark with an accuracy of 99.7% using 14 examples. As a comparison, the neural-symbolic models in the literature specialized for solving SCAN are trained with the full training set of more than 15,000 examples.

show abstract

AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Prompts

Cited by 155 publications

References 35 publications

XNLI: Explaining and Diagnosing NLI-Based Visual Data Analysis

XNLI: Explaining and Diagnosing NLI-Based Visual Data Analysis

Leveraging Large Language Models to Power Chatbots for Collecting User Self-Reported Data

Least-to-Most Prompting Enables Complex Reasoning in Large Language Models

Contact Info

Product

Resources

About