主题：1. Make LLM More Emotional: Hybrid Prompt Enhancement Training
2. Academical AI Research in the Post-GPT Era

报告人：1. Zhihang Lin, Director of Algorithm and Partner of Westlake Xinchen.Inc
2. Junbo Zhao, Professor, Zhejiang University & Former Ph.D. Student of Yann LeCun

主持人：Ruoyu Sun, Associate Professor, School of Data Science, CUHK-Shenzhen

日期：19 June (Monday), 2023

时间：10:00 to 11:20 AM, Beijing Time

地点：103 Meeting Room, Daoyuan Building

语言：English

摘要：

1. The advent of ChatGPT has made LLM (Large Language Model) one of the hottest research areas. In this talk, we will discuss our exploration of prompting for emotional AI. More specifically, we will discuss two topics:

1. How to design the training tasks so that LLM becomes "emotional" and not so robotic?

2. How we can use prompts to improve LLM's capabilities, besides controlling and influencing the LLM's output?

2. The recent boom of the large language models (LLMs) have swept the world in various aspects. Indeed, the AI research in this Post-GPT era would have to evolve and adapt. In this talk, by first recapping the development of LLMs in the past, we outline some potential terrain altering in research areas such as natural language processing. Further, we want to provide some potentially new directions, such as multimodal GPTs, data-centric tuning schemes, modular combination of large- and small-model, etc.

简介：

1. As the leader of the Rio LLM and the Westlake LLM project and participate in the development from 0-1 and continue update. Core participant of key project that sponsor by National Natural Science Foundation. As a core participant in the development of open source projects on github, the number of stars received exceeds 20K.

2. Junbo Zhao (Jake), currently employed as a PhD supervisor under ZJU100-Young researcher program at Zhejiang University. He is also a program leader of the Turing Class. Dr. Zhao received his Ph.D. degree from New York University, advised by 2018 Turing Award Laureate Yann LeCun. During his PhD, he also worked at Facebook (now Meta) AI Research Lab, and engaged in founding an autonomous driving team at NVIDIA even earlier.

Dr. Zhao has published some highly-cited papers in the Computer Science and AI venues. His papers widely cover Deep Learning, Un-, weakly- and self-supervised Learning, Large Language Models, AIGC, Data-centric Research, etc. Dr. Zhao’s papers have been cited for more than 14500 times according to Google Scholar. In addition, Dr. Zhao was elected as the cover figure of Forbes China 30 Under 30 Science and Technology Track.