Fanjia Yan

I'm a Member of Technical Staff at Microsoft AI, where I work on instruction following for coding and agents. Previously, I was an Applied Scientist at Amazon AGI, where I worked on agentic post-training for language models as a core contributor to Nova 2 Lite, spanning agentic reinforcement learning, tool-use training data, and agent evaluations.

I received my M.S. in Electrical Engineering and Computer Science and my undergraduate degrees in Computer Science and Applied Mathematics from UC Berkeley, where I worked on Language Model Agents and Tool usage. I have previously interned at Cohere, Nvidia, Amazon, and Aqueduct.

Email  /  CV  /  Scholar  /  Twitter  /  GitHub

Fanjia Yan

Research

I'm generally interested in LLM agentic frameworks and function calling / tool usage. Some of my work is listed below:

Berkeley Function Calling Leaderboard The Berkeley Function Calling Leaderboard (BFCL): From Tool Use to Agentic Evaluation of Large Language Models
Shishir G. Patil, Huanzhi Mao, Fanjia Yan, Charlie Cheng-Jie Ji, Vishnu Suresh, Ion Stoica, Joseph E. Gonzalez
ICML 2025
paper / leaderboard / code

A benchmark for evaluating function calling and agentic capabilities of language models.

LiveCodeBench
LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
Naman Jain, King Han, Alex Gu, Wen-Ding Li, Fanjia Yan, Tianjun Zhang, Sida Wang, Armando Solar-Lezama, Koushik Sen, Ion Stoica
ICLR 2025
project page / code / arXiv

LiveCodeBench collects problems from periodic contest platforms and uses them for constructing a holistic benchmark for evaluating Code LLMs across a variety of code-related scenarios continuously over time.

Gorilla OpenFunctions
Openfunctions
Charlie Cheng-Jie Ji*, Huanzhi Mao*, Fanjia Yan*, Ion Stoica, Joseph E. Gonzalez, Shishir G. Patil, Tianjun Zhang

project page / Huggingface page

A SOTA open-source, general-purpose function calling model.