Zhenting Wang

Research Scientist, MBZUAI Institute of Foundation Models

Silicon Valley Lab · Sunnyvale, California

Illustrated portrait of Zhenting Wang

I am a Research Scientist at the Institute of Foundation Models (IFM), where I work on post-training the K2 series of open-weight foundation models. I am a core contributor to K2 Horizon, a family of six models that includes the flagship 375B-A23B mixture-of-experts model.

For K2 Horizon, I ran the production reinforcement learning (RL) training and drove the design and implementation of the RL recipe and training infrastructure. I also contributed to mid-training, RL data, and evaluation, especially for agentic capabilities.

Before joining IFM, I was a senior manager-level researcher at Accenture’s Center for Advanced AI, a research scientist intern at Meta GenAI, and a research intern at Sony AI. I earned my PhD in computer science at Rutgers University.

Research interests Large language models, reinforcement learning, and agent reasoning.

Selected Publications

Full list on Google Scholar

Post-training / RL

Agent Data

Other Research