DFM: Dialogue foundation model for universal large-scale dialogue-oriented task learning

Building a universal conversational agent has been a long-standing goal of the dialogue research community. Most previous works only focus on a small set of dialogue tasks. In this work, we aim to build a unified dialogue foundation model (DFM) which can be used to solve massive diverse dialogue tas...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhi Chen, Da Ma, Hanqi Li, Lu Chen, Jiabao Ji, Yuncong Liu, Bei Chen, Mengyue Wu, Su Zhu, Xin Dong, Fujiang Ge, Qingliang Miao, Jian-Guang Lou, Shuai Fan, Kai Yu
Format: Article
Language:English
Published: KeAi Communications Co. Ltd. 2025-01-01
Series:AI Open
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2666651025000075
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849390541840580608
author Zhi Chen
Da Ma
Hanqi Li
Lu Chen
Jiabao Ji
Yuncong Liu
Bei Chen
Mengyue Wu
Su Zhu
Xin Dong
Fujiang Ge
Qingliang Miao
Jian-Guang Lou
Shuai Fan
Kai Yu
author_facet Zhi Chen
Da Ma
Hanqi Li
Lu Chen
Jiabao Ji
Yuncong Liu
Bei Chen
Mengyue Wu
Su Zhu
Xin Dong
Fujiang Ge
Qingliang Miao
Jian-Guang Lou
Shuai Fan
Kai Yu
author_sort Zhi Chen
collection DOAJ
description Building a universal conversational agent has been a long-standing goal of the dialogue research community. Most previous works only focus on a small set of dialogue tasks. In this work, we aim to build a unified dialogue foundation model (DFM) which can be used to solve massive diverse dialogue tasks. To achieve this goal, a large-scale well-annotated dialogue dataset with rich task diversity (DialogZoo) is collected. We introduce a framework to unify all dialogue tasks and propose novel auxiliary self-supervised tasks to achieve stable training of DFM on the highly diverse large scale DialogZoo corpus. Experiments show that, compared with models of the same size, DFM can achieve competitive performance on very rich cross-domain downstream dialogue tasks. Furthermore, when scaling to large language models, DFM remains effective. This demonstrates that DFM largely extends the ability of unified dialogue pre-trained model.
format Article
id doaj-art-6ff577a130624a829f4c1f5251b61916
institution Kabale University
issn 2666-6510
language English
publishDate 2025-01-01
publisher KeAi Communications Co. Ltd.
record_format Article
series AI Open
spelling doaj-art-6ff577a130624a829f4c1f5251b619162025-08-20T03:41:34ZengKeAi Communications Co. Ltd.AI Open2666-65102025-01-01610811710.1016/j.aiopen.2025.04.001DFM: Dialogue foundation model for universal large-scale dialogue-oriented task learningZhi Chen0Da Ma1Hanqi Li2Lu Chen3Jiabao Ji4Yuncong Liu5Bei Chen6Mengyue Wu7Su Zhu8Xin Dong9Fujiang Ge10Qingliang Miao11Jian-Guang Lou12Shuai Fan13Kai Yu14X-LANCE Lab, School of Computer Science MoE Key Lab of Artificial Intelligence, SJTU AI Institute, Shanghai Jiao Tong University, Shanghai, ChinaX-LANCE Lab, School of Computer Science MoE Key Lab of Artificial Intelligence, SJTU AI Institute, Shanghai Jiao Tong University, Shanghai, ChinaX-LANCE Lab, School of Computer Science MoE Key Lab of Artificial Intelligence, SJTU AI Institute, Shanghai Jiao Tong University, Shanghai, ChinaX-LANCE Lab, School of Computer Science MoE Key Lab of Artificial Intelligence, SJTU AI Institute, Shanghai Jiao Tong University, Shanghai, China; Corresponding authors.X-LANCE Lab, School of Computer Science MoE Key Lab of Artificial Intelligence, SJTU AI Institute, Shanghai Jiao Tong University, Shanghai, ChinaX-LANCE Lab, School of Computer Science MoE Key Lab of Artificial Intelligence, SJTU AI Institute, Shanghai Jiao Tong University, Shanghai, ChinaMicrosoft Research Asia, Beijing, ChinaX-LANCE Lab, School of Computer Science MoE Key Lab of Artificial Intelligence, SJTU AI Institute, Shanghai Jiao Tong University, Shanghai, ChinaAISpeech Co., Ltd., Suzhou, ChinaAISpeech Co., Ltd., Suzhou, ChinaAISpeech Co., Ltd., Suzhou, ChinaAISpeech Co., Ltd., Suzhou, ChinaMicrosoft Research Asia, Beijing, ChinaAISpeech Co., Ltd., Suzhou, ChinaX-LANCE Lab, School of Computer Science MoE Key Lab of Artificial Intelligence, SJTU AI Institute, Shanghai Jiao Tong University, Shanghai, China; Corresponding authors.Building a universal conversational agent has been a long-standing goal of the dialogue research community. Most previous works only focus on a small set of dialogue tasks. In this work, we aim to build a unified dialogue foundation model (DFM) which can be used to solve massive diverse dialogue tasks. To achieve this goal, a large-scale well-annotated dialogue dataset with rich task diversity (DialogZoo) is collected. We introduce a framework to unify all dialogue tasks and propose novel auxiliary self-supervised tasks to achieve stable training of DFM on the highly diverse large scale DialogZoo corpus. Experiments show that, compared with models of the same size, DFM can achieve competitive performance on very rich cross-domain downstream dialogue tasks. Furthermore, when scaling to large language models, DFM remains effective. This demonstrates that DFM largely extends the ability of unified dialogue pre-trained model.http://www.sciencedirect.com/science/article/pii/S2666651025000075Dialogue foundation modelPre-trained language modelMultitask learningKnowledge-grounded dialogue
spellingShingle Zhi Chen
Da Ma
Hanqi Li
Lu Chen
Jiabao Ji
Yuncong Liu
Bei Chen
Mengyue Wu
Su Zhu
Xin Dong
Fujiang Ge
Qingliang Miao
Jian-Guang Lou
Shuai Fan
Kai Yu
DFM: Dialogue foundation model for universal large-scale dialogue-oriented task learning
AI Open
Dialogue foundation model
Pre-trained language model
Multitask learning
Knowledge-grounded dialogue
title DFM: Dialogue foundation model for universal large-scale dialogue-oriented task learning
title_full DFM: Dialogue foundation model for universal large-scale dialogue-oriented task learning
title_fullStr DFM: Dialogue foundation model for universal large-scale dialogue-oriented task learning
title_full_unstemmed DFM: Dialogue foundation model for universal large-scale dialogue-oriented task learning
title_short DFM: Dialogue foundation model for universal large-scale dialogue-oriented task learning
title_sort dfm dialogue foundation model for universal large scale dialogue oriented task learning
topic Dialogue foundation model
Pre-trained language model
Multitask learning
Knowledge-grounded dialogue
url http://www.sciencedirect.com/science/article/pii/S2666651025000075
work_keys_str_mv AT zhichen dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT dama dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT hanqili dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT luchen dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT jiabaoji dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT yuncongliu dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT beichen dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT mengyuewu dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT suzhu dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT xindong dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT fujiangge dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT qingliangmiao dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT jianguanglou dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT shuaifan dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning
AT kaiyu dfmdialoguefoundationmodelforuniversallargescaledialogueorientedtasklearning