Hyung Won Chung
Researcher on reasoning and agents, formerly at Google Brain; first author of the Flan-T5 scaling paper
Hyung Won Chung's homepage describes research on reasoning and agents at OpenAI, including Foundational Contributor credits on o1-preview, o1, and Deep Research and leading Codex mini model training. Before OpenAI, Chung was at Google Brain, building the JAX-based large-scale training system T5X, training large models such as PaLM, and working on instruction fine-tuning for the Flan-PaLM and Flan-T5 model families. Chung is first author of Scaling Instruction-Finetuned Language Models, which publicly released Flan-T5 checkpoints; the Flan-T5 XXL model card on Hugging Face lists the Apache 2.0 license.123
Work
Flan-T5 (external site: huggingface.co)Open weights
First author of Scaling Instruction-Finetuned Language Models (2022), which studies instruction finetuning across tasks, model sizes, and chain-of-thought data and publicly releases Flan-T5 checkpoints; the Flan-T5 XXL model card cites the paper and lists the Apache 2.0 license.23
Earlier roles
- Research Scientist, OpenAI1
- Researcher (T5X, PaLM, Flan-PaLM and Flan-T5), Google Brain1
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.