Publications

ChatGPT’s One-year Anniversary: Are Open-Source Large Language Models Catching up?
Benchmarking open-source LLMs outperforming ChatGPT.
Hailin Chen, Fangkai Jiao, Chengwei Qin, Xingxuan Li, Mathieu Ravaut, Ruochen Zhao, Caiming Xiong, Shafiq Joty
PDF Cite Code
ChatGPT’s One-year Anniversary: Are Open-Source Large Language Models Catching up?
On Context Utilization in Summarization with Large Language Models
Analyzing how several open-source LLMs process their context in abstractive summarization tasks.
Mathieu Ravaut, Shafiq Joty, Aixin Sun, Nancy F. Chen
PDF Cite
On Context Utilization in Summarization with Large Language Models
PromptSum: Parameter-Efficient Controllable Abstractive Summarization
A method to build summarization models which are high-performing with small supervision and few changed parameters, as well as controllable.
Mathieu Ravaut, Hailin Chen, Ruochen Zhao, Chengwei Qin, Shafiq Joty, Nancy F. Chen
PDF Cite Code
PromptSum: Parameter-Efficient Controllable Abstractive Summarization
Unsupervised Summarization Re-ranking
Second-stage approach for abstractive summarization consisting in re-ranking summary candidates in an unsupervised manner. It also extends into a self-training objective.
Mathieu Ravaut, Shafiq Joty, Nancy Chen
PDF Cite Code Poster Slides Source Document
Unsupervised Summarization Re-ranking
A Data-centric Framework for Improving Domain-specific Machine Reading Comprehension Datasets
Compaing data enhancement techniques on machine reading comprehension datasets.
Iva Bojic, Josef Halim, Verena Suharman, Sreeja Tar, Qi Chwen Ong, Duy Phung, Mathieu Ravaut, Shafiq Joty, Josip Car
PDF Cite Code Source Document
A Data-centric Framework for Improving Domain-specific Machine Reading Comprehension Datasets
Towards Summary Candidates Fusion
New second-stage approach to abstractive summarization, consisting in fusing together existing summary candidates through a Fusion-In-Decoder approach.
Mathieu Ravaut, Shafiq Joty, Nancy Chen
PDF Cite Code Poster Slides Source Document
Towards Summary Candidates Fusion
SummaReranker: A Multi-Task Mixture-of-Expert Re-ranking Framework for Abstractive Summarization
Second-stage approach to abstractive summarization, re-ranking candidates obtained trough beam search, diverse beam search, etc by jointly optimizing on several metrics (e.g, ROUGE-1, BERTScore).
Mathieu Ravaut, Shafiq Joty, Nancy Chen
PDF Cite Code Poster Slides Source Document
SummaReranker: A Multi-Task Mixture-of-Expert Re-ranking Framework for Abstractive Summarization
Predicting Twitter Engagement With Deep Language Models
Usage of large language models embeddings to improve recommendation performance on large-scale Twitter dataset.
Maksims Volkovs, Zhaoyue Cheng, Mathieu Ravaut, Hojin Yang, Kevin Shen, Jin Peng Zhou, Anson Wong, Saba Zuberi, Ivan Zhang, Nick Frosst, Helen Ngo, Carol Chen, Bharat Venkitesh, Stephen Gou, Aidan N. Gomez
Cite Code Source Document
Predicting Twitter Engagement With Deep Language Models
Truly Multi-modal YouTube-8M Video Classification with Video, Audio, and Text
Leveraging additional text features on top of audio and visual ones to improve classification of YouTube videos.
Zhe Wang, Kingsley Kuan, Mathieu Ravaut, Gaurav Manek, Sibo Song, Yuan Fang, Seokhwan Kim, Nancy F. Chen, Luis Fernandon D'Haro, Luu Anh Tuan, Hongyang Zhu, Zeng Zeng, Ngai Man Cheung, Georgios Piliouras, Jie Lin, Vijay Chandrasekhar
PDF Cite Code Source Document
Truly Multi-modal YouTube-8M Video Classification with Video, Audio, and Text
Deep Learning Applied to Underwater Mine Warfare
A deep learning approach to detect mines in the seabed from Sonar data.
Killian Denos, Mathieu Ravaut, Antoine Fagette, Hock-Siong Lim
Cite Source Document
Deep Learning Applied to Underwater Mine Warfare