Jonathan May

Jonathan May is a Research Associate Professor in the Department of Computer Science at the University of Southern California and a Principal Scientist at the Information Sciences Institute. He serves as Director of the Center for Useful Techniques Enhancing Language Applications Based on Natural And Meaningful Evidence (CUTELABNAME). Dr. May teaches courses including CSCI 544: Applied Natural Language Processing and CSCI 662: Advanced Natural Language Processing through Fall 2024.

Dr. May's research spans natural language processing, machine translation, and computational linguistics with significant contributions to cross-lingual processing, dialogue systems, and language model development. His work bridges theoretical advances with practical applications in journalism, negotiation systems, and multimodal understanding. Recent research focuses on large language models, with particular emphasis on efficiency, personalization, grounded generation, and exploring the limits of what language models can understand about physical phenomena and human communication.

Analysis of Dr. May's recent publications (2024-2025) reveals a strong focus on advancing large language models, with particular emphasis on efficiency (e.g., 'Megalodon: Efficient LLM Pretraining and Inference with Unlimited Context Length'), personalization (e.g., 'Tuning-Free Personalized Alignment via Trial-Error-Explain In-Context Learning'), and grounded generation (e.g., 'NewsInterview: a Dataset and a Playground to Evaluate LLMs' Ground Gap via Informational Interviews'). A notable trend is the integration of reflection and memory mechanisms into language models, as well as exploring multimodal understanding across domains from food science to negotiation.

• Outstanding Paper Award at EMNLP 2024 for 'Are Large Language Models Capable of Generating Human-Level Narratives?'
• Outstanding Paper Award at NAACL 2018 for 'Recurrent Neural Networks as Weighted Language Recognizers'
• Best Demo Award at ACL Demo Sessions 2018 for 'Out-of-the-box Universal Romanization Tool uroman'

Dr. May leads CUTELABNAME, which has produced numerous software tools including uroman (Universal Romanizer), BotEval, and SPOLIN. The lab maintains an active research agenda with regular publications at top NLP conferences and collaborations across academia and industry. His work demonstrates both theoretical rigor and practical applicability, with many techniques being adopted in real-world NLP systems.