Large Language Models are Zero-Shot Reasoners

Kojima T., Gu S. S., Reid M., Matsuo Y. and Iwasawa Y. Large language models are zero-shot reasoners. NIPS, 2022.

概

Zero-shot CoT.

Let's think step by step

• 如上图所示, 利用 "Let's think step by step" 来引导推理, 最后利用 "Therefore, the answer (arabic numerals) is" 来引出最后的答案.

• Zero-shot 引导词还有如下的选择:
  

代码

[official]

本栏目推荐文章
• 【略读论文|大模型相关】Zero-Shot Relational Learning on Temporal Knowledge Graphs with Large Language Models
• 南阳师范大学 外国语学院 Foreign language school
• 413 Request Entity Too Large
• SciTech-BigDataAIML-Tensorflow-Introduction to modules, layers, and models
• 大模型评测-微软亚洲研究院：A Survey on Evaluation of Large Language Models论文分享
• 【五期李伟平】CCF-B（TFS'23）Consensus Reaching Process With Multiobjective Optimization for Large-Scale Group Decision Making With Cooperative Game
• Language Models are Unsupervised Multitask Learners
• InstructGPT《InstructGPT: Training language models to follow instructions with human feedback》解读
• [Nginx] API请求体过大时，报：Nginx 413 Request Entity Too Large [转载]
• GPT-3《Language Models are Few-Shot Learners》解读