| |
Training Methods for Retrieval-Augmented Generative Question Answering
|
Speaker (s):
Alessandro Moschitti
Principal Researcher
Amazon Alexa AI
|
|
Date:
Time:
Venue:
|
|
5 December 2023, Tuesday
11:00am – 12:00pm
School of Economics/School of Computing
& Information Systems 2 (SOE/SCIS 2)
Level 4, Seminar Room 4-3
Singapore Management University
90 Stamford Road, Singapore 178903
Please register by 1 December 2023.
We look forward to seeing you at this research seminar.
|
|
About the Talk
Reinforcement Learning from Human Feedback (RLHF) is a main aspect of Large Language Model (LLM) research, which has contributed to obtain the breakthrough in answer generation. Retrieval-Augmented Generation (RAG) has enabled LLMs to use recent knowledge, thus allowing them to answer questions about latest information, and reducing hallucinations, at the same time. In this talk, we present a different viewpoint of RAG-LLMs based on our experience in building Generative Question Answering (QA) systems. We designed generative systems as a new approach to joint-modeling the answer candidates selected by accurate passage re-rankers. Moreover, we use the latter as reward functions to improve fine-tuning of the generative models. Our approach provides a more direct interpretation of RAG-LLMs, reward functions, and knowledge grounding. This view also inspired the designing of our latest techniques for RLHF based on automatic QA evaluators.
About the Speaker
Alessandro Moschitti is a Principal Research Scientist of Amazon Alexa AI, where he has been leading the science of Alexa information service since 2018. He designed the Alexa QA system based on unstructured text and more recently the first Generative QA system to extend the answer skills of Alexa. He obtained his Ph.D. in CS from the University of Rome in 2003, and then did his postdoc at The University of Texas at Dallas for two years. He was professor of the CS Dept. of the University of Trento, Italy, from 2007 to 2021. He participated to the Jeopardy! Grand Challenge with the IBM Watson Research center (2009 to 2011), and collaborated with them until 2015. He was a Principal Scientist of the Qatar Computing Research Institute (QCRI) for five years (2013-2018). His expertise concerns theoretical and applied machine learning in the areas of NLP, IR and Data Mining. He is well-known for his work on structural kernels and neural networks for syntactic/semantic inference over text, documented by more than 330 scientific articles. He has received four IBM Faculty Awards, one Google Faculty Award, and five best paper awards. He was the General Chair of EACL 2023 and EMNLP 2014, a PC co-Chair of CoNLL 2015, and has had a chair role in more than 70 conferences and workshops. He is currently a senior action/associate editor of ACM Computing Survey and JAIR. He has led ~30 research projects, e.g., with MIT CSAIL
|
