Advances in reasoning by prompting large language models: A survey
Authors:
Open Access
- 1 March 2026
- journal article
- Published by Tsinghua University Press in Cybernetics and Intelligence
- Vol. 1 (1) , 9390004
Abstract
Reasoning is not only an essential aspect of human intelligence but also one of the main research topics in artificial intelligence. With the recent revolutionary developments in natural language processing, it has been observed that large language models possess a degree of reasoning capabilities. When elicited by prompting, these models can exhibit impressive performance in various reasoning tasks. In this paper, we survey the recent advances in reasoning by prompting large language models. We provide an overview of key benchmarks and categorize the different reasoning methods. Our survey focuses on the most recent advancements in this field and seeks to provide a comprehensive understanding of the current state-of-the-art.Keywords
This publication has 59 references indexed in Scilit:
- Shortcut Learning of Large Language Models in Natural Language UnderstandingCommunications of the ACM, 2023
- Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language ProcessingACM Computing Surveys, 2023
- Logic Tensor NetworksArtificial Intelligence, 2022
- Neural probabilistic logic programming in DeepProbLogArtificial Intelligence, 2021
- From machine learning to machine reasoningMachine Learning, 2013
- Ensembling neural networks: Many could be better than allArtificial Intelligence, 2002
- Learning to reasonJournal of the ACM, 1997
- Why Expert Systems FailJournal of the Operational Research Society, 1985
- Reasoning about a RuleQuarterly Journal of Experimental Psychology, 1968
- The logic theory machine--A complex information processing systemIEEE Transactions on Information Theory, 1956