Jonathan Pei
Under-graduate student @ UC Berkeley, Berkeley NLP Group
controlled text generation
conversational ai
question generation
personalization
prompting
open-domain dialogue systems
toxicity mitigation
bias reduction
decoding-time methods
sentiment control
personalized question generation
1
presentations
SHORT BIO
Jonathan is an undergraduate at UC Berkeley studying EECS and Statistics, and performs research with the Berkeley NLP group. He is interested in controlled text generation, long-text question answering, and symbolic reasoning for deep learning applications. Beyond NLP or PL, he likes learning about statistics, probability theory, signal processing, and various physics topics.
Presentations
PREADD: Prefix-Adaptive Decoding for Controlled Text Generation
Jonathan Pei and 2 other authors