Yulu Gan

Hi, I am Yulu Gan, a third-year CS PhD at MIT, studying AI and Science. Advised by Tomaso Poggio and Phillip Isola. My research focuses on LLM Post-Training, LLM Agents and Agentic Robotics. I have interned at Microsoft Research Asia (MSRA), and NVIDIA Research (GEAR lab).

Recent Publications* indicates equal contribution.

All publications →

Neural Thickets: Diverse Task Experts Are Dense Around Pretrained WeightsICML 2026 Spotlight

Yulu Gan, Phillip Isola

Paper Project tl;dr 640
Small model Large model pretrained weights Gaussian search window USPTO chemistry ROCStories writing MBPP code GSM8k math
Notebook preview for Neural Thickets blog A Game of Random Guessing Blog Open notebook Video preview for Pretrained LLMs Are Surrounded by Task Experts Pretrained LLMs Are Surrounded by Task Experts feat Yulu Gan from MIT 2,465 views 107 likes 3 comments

Evolution Strategies at Scale: LLM Fine-Tuning Beyond Reinforcement LearningICML 2026

Xin Qiu*, Yulu Gan*, Conor F. Hayes*, Qiyao Liang, Yinggan Xu, Roberto Dailey, Elliot Meyerson, Babak Hodjat, Risto Miikkulainen

Paper tl;dr 381
Reward landscape current weights sampled perturbations high reward
Video preview for A new way to fine-tune LLMs just dropped A new way to fine-tune LLMs just dropped 48,640 views 2,620 likes 178 comments Video preview for How to Finetune 8 Billion Parameters with Evolution Strategies? How to Finetune 8 Billion Parameters with Evolution Strategies? 3,248 views 151 likes 12 comments Video preview for exploring Evolution Strategies at Scale for LLM Finetuning exploring "Evolution Strategies at Scale for LLM Finetuning" | Deep Learning Study Session 1,168 views 45 likes 11 comments

Selected Awards

McGovern Institute Student Fellow
Dr. Chiang Chen Fellowship
Frederick R. (1953) and Barbara Cronin Fellowship
AAAI Best Student Paper Award (project lead)
Second Place in RoboCup China Open (team lead)
First Place in China Mathematical Olympiad, Zhejiang Division

Blog

See all →
To be updated.

Reading & Talks

See all →