Is ChatGPT a General-Purpose Natural Language Processing Task Solver?

Является ли ChatGPT универсальным решателем задач обработки естественного языка?
Michihiro Yasunaga, Chengwei Qin, Zhuosheng Zhang, Diyi Yang, Aston Zhang, Jiaao Chen
2023-01-01

ChatGPTarithmetic reasoningnatural language processing taskssequence taggingzero-shot learning
Spurred by advancements in scale, large language models (LLMs) have demonstrated the ability to perform a variety of natural language processing (NLP) tasks zero-shot-i.e., without adaptation on downstream data.Recently, the debut of ChatGPT 1 has drawn a great deal of attention from the natural language processing (NLP) community due to the fact that it can generate high-quality responses to human input and self-correct previous mistakes based on subsequent conversations.However, it is not yet known whether ChatGPT can serve as a generalist model that can perform many NLP tasks zero-shot.In this work, we empirically analyze the zero-shot learning ability of ChatGPT by evaluating it on 20 popular NLP datasets covering 7 representative task categories.With extensive empirical studies, we demonstrate both the effectiveness and limitations of the current version of ChatGPT.We find that ChatGPT performs well on many tasks favoring reasoning capabilities (e.g., arithmetic reasoning) while it still faces challenges when solving specific tasks such as sequence tagging.We additionally provide in-depth analysis through qualitative case studies.
1
ChatGPT performs effectively on many NLP tasks that favor reasoning capabilities, including arithmetic reasoning.
2
ChatGPT remains limited on task-specific problems such as sequence tagging, indicating it is not uniformly general-purpose.
3
Qualitative case studies provide deeper analysis of ChatGPT’s zero-shot successes and failure modes.
4
The study evaluates ChatGPT’s zero-shot performance on 20 popular NLP datasets spanning seven representative task categories.

ChatGPT as a large language model evaluated on 20 NLP datasets spanning 7 task categories

ChatGPT’s zero-shot performance, effectiveness, and limitations across diverse NLP tasks, including reasoning and sequence tagging

Publication Details
Publication Date
2023-01-01
Journal
Publisher
ISSN
Cited by
888
Access Type
Author Information
Authors
Michihiro Yasunaga
Chengwei Qin
Zhuosheng Zhang
Diyi Yang
Aston Zhang
Jiaao Chen
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%