We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. The empirical gains can be striking. For instance, prompting a 540B-parameter language model with just eight chain of thought exemplars achieves state of the art accuracy on the GSM8K benchmark of math word problems, surpassing even finetuned GPT-3 with a verifier.
来源出处
Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
http://arxiv.org/abs/2201.11903
相关内容
发布日期
08/04/2020 - 01:35
发布日期
06/17/2022 - 10:21
发布日期
06/22/2024 - 17:53
发布日期
01/10/2022 - 19:32
发布日期
09/21/2023 - 22:52
发布日期
02/10/2022 - 15:24
发布日期
01/10/2022 - 19:31
发布日期
08/04/2020 - 01:35
发布日期
10/23/2024 - 19:39
发布日期
09/02/2024 - 19:26
发布日期
08/04/2020 - 01:35
发布日期
08/04/2020 - 01:35
发布日期
06/07/2024 - 17:46
发布日期
08/20/2024 - 19:21
发布日期
10/31/2021 - 01:12
发布日期
04/18/2024 - 09:29
发布日期
08/04/2020 - 01:35
发布日期
09/18/2024 - 19:30
发布日期
09/02/2024 - 19:26
发布日期
07/02/2023 - 18:27