Knowledge Commons of Institute of Automation,CAS
Generative Calibration for In-context Learning | |
Zhongtao Jiang1,2; Yuanzhe Zhang1,2; Cao Liu3; Jun Zhao1,2; Kang Liu1,2,4 | |
2023-10-06 | |
会议名称 | Findings of the Association for Computational Linguistics: EMNLP 2023 |
会议日期 | 2023-10-6 |
会议地点 | Singapore |
摘要 | As one of the most exciting features of large language models (LLMs), in-context learning is a mixed blessing. While it allows users to fast-prototype a task solver with only a few training examples, the performance is generally sensitive to various configurations of the prompt such as the choice or order of the training examples. In this paper, we for the first time theoretically and empirically identify that such a paradox is mainly due to the label shift of the in-context model to the data distribution, in which LLMs shift the label marginal p(y) while having a good label conditional p(x|y). With this understanding, we can simply calibrate the in-context predictive distribution by adjusting the label marginal, which is estimated via Monte-Carlo sampling over the in-context model, i.e., generation of LLMs. We call our approach as generative calibration. We conduct exhaustive experiments with 12 text classification tasks and 12 LLMs scaling from 774M to 33B, generally find that the proposed method greatly and consistently outperforms the ICL as well as state-of-the-art calibration methods, by up to 27% absolute in macro-F1. Meanwhile, the proposed method is also stable under different prompt configurations. |
七大方向——子方向分类 | 自然语言处理 |
国重实验室规划方向分类 | 语音语言处理 |
是否有论文关联数据集需要存交 | 否 |
文献类型 | 会议论文 |
条目标识符 | http://ir.ia.ac.cn/handle/173211/57263 |
专题 | 复杂系统认知与决策实验室 |
作者单位 | 1.The Laboratory of Cognition and Decision Intelligence for Complex Systems, Institute of Automation, Chinese Academy of Sciences 2.School of Artificial Intelligence, University of Chinese Academy of Sciences 3.Meituan 4.Shanghai Artificial Intelligence Laboratory |
第一作者单位 | 中国科学院自动化研究所 |
推荐引用方式 GB/T 7714 | Zhongtao Jiang,Yuanzhe Zhang,Cao Liu,et al. Generative Calibration for In-context Learning[C],2023. |
条目包含的文件 | 下载所有文件 | |||||
文件名称/大小 | 文献类型 | 版本类型 | 开放类型 | 使用许可 | ||
Generative Calibrati(763KB) | 会议论文 | 开放获取 | CC BY-NC-SA | 浏览 下载 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论