当前位置:网站首页>Naacl-22 | introduce the setting of migration learning on the prompt based text generation task
Naacl-22 | introduce the setting of migration learning on the prompt based text generation task
2022-07-04 22:32:00 【Zhiyuan community】
The pre training language model has made significant progress in text generation tasks through fine-tuning , But in the scenario of sparse data , It is usually impossible to fine tune directly . therefore , In this paper, based on prompt Setting of transfer learning . The author first learns one for different tasks in the source domain prompt, Thus construct prompt pool , Then migrate in the target task . In order to consider both task level and instance level information , The author designed an adaptive attention mechanism , For each instance sample in the target task , The model will select the most relevant source task for it prompt. The author has carried out experiments on various generation tasks and data sets , The results show that the migration method proposed by the author can improve the generation effect on the target task very well .
Paper title :
Learning to Transfer Prompts for Text Generation
Thesis link :
https://arxiv.org/abs/2205.01543
边栏推荐
- La prospérité est épuisée, les choses sont bonnes et mauvaises: Où puis - je aller pour un chef de station personnel?
- i. Mx6ull driver development | 24 - platform based driver model lights LED
- 关于栈区、堆区、全局区、文字常量区、程序代码区
- Solana chain application crema was shut down due to hacker attacks
- PHP short video source code, thumb animation will float when you like it
- PostgreSQL服务端编程聚合和分组
- # 2156. Find the substring of the given hash value - post order traversal
- Tla+ introductory tutorial (1): introduction to formal methods
- Scala download and configuration
- 卷积神经网络模型之——LeNet网络结构与代码实现
猜你喜欢
抖音实战~评论数量同步更新
Visual task scheduling & drag and drop | scalph data integration based on Apache seatunnel
Radio and television Wuzhou signed a cooperation agreement with Huawei to jointly promote the sustainable development of shengteng AI industry
醒悟的日子,我是怎么一步一步走向软件测试的道路
2022-07-04:以下go语言代码输出什么?A:true;B:false;C:编译错误。 package main import “fmt“ func main() { fmt.Pri
Machine learning notes mutual information
UML diagram memory skills
LOGO special training camp section I identification logo and Logo Design Ideas
A large number of virtual anchors in station B were collectively forced to refund: revenue evaporated, but they still owe station B; Jobs was posthumously awarded the U.S. presidential medal of freedo
Use blocconsumer to build responsive components and monitor status at the same time
随机推荐
Play with grpc - go deep into concepts and principles
How can the advertising system of large factories be upgraded without the presence of large models
赋能数字经济 福昕软件出席金砖国家可持续发展高层论坛
Concurrent network modular reading notes transfer
High school physics: linear motion
常用的开源无代码测试工具
PostgreSQL JOIN实践及原理
Logo Camp d'entraînement section 3 techniques créatives initiales
Solana chain application crema was shut down due to hacker attacks
Éducation à la transmission du savoir | Comment passer à un test logiciel pour l'un des postes les mieux rémunérés sur Internet? (joindre la Feuille de route pour l'apprentissage des tests logiciels)
MySQL存储数据加密
Mysql root 账号如何重置密码
Redis sentinel simply looks at the trade-offs between distributed high availability and consistency
湘江鲲鹏加入昇腾万里伙伴计划,与华为续写合作新篇章
LOGO特训营 第四节 字体设计的重要性
LOGO特訓營 第一節 鑒別Logo與Logo設計思路
[Yugong series] go teaching course 003-ide installation and basic use in July 2022
MySQL storage data encryption
Easy to use app recommendation: scan QR code, scan barcode and view history
La prospérité est épuisée, les choses sont bonnes et mauvaises: Où puis - je aller pour un chef de station personnel?