# The ADAIO System at the BEA-2023 Shared Task on Generating AI Teacher Responses in Educational Dialogues

**Authors:** Adaeze Adigwe, Zheng Yuan  
**arXiv:** [2306.05360](https://arxiv.org/abs/2306.05360)  
**Date:** 2023-06-08  
**Summary page:** https://www.summarizepaper.com/en/paper/2306.05360v1/

## Key points

- ADAIO team's system entry in the BEA 2023 Shared Task on Generating AI Teacher Responses
- Development of a system involving assessment of baseline models using OpenAI GPT-3
- Designing diverse prompts to prompt OpenAI models for teacher response generation
- Achieving second place in the challenge with a few-shot prompt-based approach using OpenAI text-davinci-003 model
- Demonstrating the few-shot learning capabilities of large-language models, particularly OpenAI's GPT-3, as AI teachers
- Utilizing fine-tuning strategies to enhance the performance of their system in generating suitable responses during educational dialogues
- Contribution to advancing education by demonstrating how state-of-the-art generative models can effectively serve as AI teachers
- Potential for improving and enhancing student learning experiences

## Summary

This paper presents the ADAIO team's system entry in the Building Educational Applications (BEA) 2023 Shared Task on Generating AI Teacher Responses in Educational Dialogues. The task aimed to evaluate the performance of state-of-the-art generative models as AI teachers in generating appropriate responses within a student-teacher dialogue. The authors developed a system that involved assessing various baseline models using OpenAI GPT-3 and designing diverse prompts to prompt the OpenAI models for teacher response generation. After participating in the challenge, the ADAIO system achieved second place by employing a few-shot prompt-based approach with the OpenAI text-davinci-003 model. This approach demonstrated the few-shot learning capabilities of large-language models, particularly OpenAI's GPT-3, when used as AI teachers. and fine-tuning strategies were utilized to enhance the performance of their system in generating suitable responses during educational dialogues. Overall, this work contributes to advancing  by demonstrating how state-of-the-art generative models can effectively serve as AI teachers. The findings suggest that these models have significant potential for improving  and enhancing student learning experiences.