Yang Chen

Building on prior work, I contributed to research combining large-scale supervised fine-tuning with RL. We released the SOTA 7B model -- AceReason-Nemotron-1.1-7B.

I co-led the effort to scale the previous work to both math and code domains. We released the SOTA medium size model -- AceReason-Nemotron-14B.

I led the research on scaling RL to train LLMs to solve math problems. We trained and released the AceMath-RL-Nemotron-7B model.

I co-led the research on developing math reasoning and reward models. We released the AceMath-1.5B, 7B, and 72B models along with AceMath-RM-7B and 72B, which surpass GPT-4o and Qwen2.5-Math at the time.