Research
Currently, I’m interested in developing new methods of modeling heterogeneous rewards through dense reward shaping or defining new, more grounded reward functions for language model tuning. Specifically, I’m invested in defining and developing a more informative reward space from sparse signals to better align AI systems to human behavior or preferences.
Dynamic Multi-Reward Weighting for Multi-Style Controllable Generation
Karin De Langis, Ryan Koo, Dongyeop Kang
Empirical Methods in Natural Language Processing (EMNLP) Main, 2024
Benchmarking Cognitive Biases in Large Language Models as Evaluators
Ryan Koo, Minhwa Lee, Vipul Raheja, Jong Inn Park, Zae Myung Kim, Dongyeop Kang
Association for Computational Linguistics (ACL) Findings, 2024
Meta-Crafting: Improved Detection of Out-of-Distributed Texts via Crafting Metadata Space
Ryan Koo, Yekyung Kim, Dongyeop Kang, Jaehyung Kim
Association for the Advancement of Artificial Intelligence (AAAI) Student Abstract, 2024
CoEdIT: Text Editing by Task-Specific Instruction Tuning
Vipul Raheja, Dhruv Kumar, Ryan Koo, Dongyeop Kang
Empirical Methods in Natural Language Processing (EMNLP) Findings, 2023