Gaokai Zhang

M.S. in Intelligent Information Systems

Carnegie Mellon University

Language Technologies Institute

Hi there! I’m Gaokai Zhang, an M.S. student in Intelligent Information Systems at CMU LTI since Fall 2025. I hold dual B.S. degrees from ZJUI (CompE @ UIUC, ECE @ ZJU).

Currently, I’m working on SWE-Bench related code-generation agent training with Kexun Zhang in Prof. Lei Li’s lab, focusing on supervised fine-tuning and reinforcement learning for coding agents. I am also working with Yiqing Xie and Prof. Daniel Fried on code-generation data synthesis and agent training.

Open to LLM-related MLE/RS opportunities as I’m graduating in December 2026!

Experience

Microsoft Research Asia (Jul 2024 - Jul 2025) Research Intern, Systems & Networking Group Mentored by Dr. Li Lyna Zhang

Led LoongRL: Novel data synthesis + reinforcement learning enabling 7B models to surpass 32B LRMs in long-context reasoning (100k-200k tokens) (ICLR 2026 Oral)
Contributed to LongRoPE2: Extended LLM context windows to 128K tokens while retaining 98.5% short-context accuracy (ICML 2025 poster)
Built parallel pipeline for large-scale user-query processing; delivered production-ready long-context recommendation models to Microsoft Asia-Pacific R&D

Carnegie Mellon University (Oct 2025 – Present) Research Assistant, Language Technologies Institute

Synthetic task generation for training coding agents to generalize across repository-level environments (Hybrid-Gym)
Synthetic issue generation and test-time training for improving LLM-based coding agents under sparse rewards

University of Illinois Urbana-Champaign Research Assistant with Prof. Fan Lai and Prof. Minjia Zhang

Monte-Carlo-Tree-Search planning for cost-efficient LLM training on heterogeneous GPUs/TPUs
Robustness benchmarking of LLMs (Stochastic Monkeys)

Research Interests

Long-context reasoning & scaling
Reinforcement learning for LLMs
Efficient training architectures
Code generation agents

Beyond Research

Outside work, I enjoy gaming (lifetime Faker fan), vibe to rap, and hunt for the perfect omakase bite.

Feel free to reach out - always happy to connect with like-minded friends and collaborators!

news

Jan 26, 2026	LoongRL accepted as Oral at ICLR 2026 - my first co-first-authored paper!
Aug 01, 2025	Started M.S. in Intelligent Information Systems at CMU LTI!
May 01, 2025	LongRoPE2 accepted at ICML 2025 as a poster presentation!
Jul 01, 2024	Joined Microsoft Research Asia as a research intern in the Systems & Networking Group.

selected publications

ICLR 2026 Oral

LoongRL: Incentivizing Long-Context Reasoning in Large Language Models via Reinforcement Learning

Siyuan Wang, Gaokai Zhang, Li Lyna Zhang, Ning Shang, Fan Yang, Dongyao Chen, and 1 more author

International Conference on Learning Representations, 2026

Abs arXiv HTML Website

We present LoongRL, a reinforcement learning framework with novel data synthesis that enables 7B parameter models to surpass 32B long-range models on long-context reasoning tasks at 100k-200k tokens.
ICML 2025

LongRoPE2: Near-Lossless LLM Context Window Scaling

Ning Shang, Li Lyna Zhang, Siyuan Wang, Gaokai Zhang, Gilsinia Lopez, Fan Yang, and 2 more authors

International Conference on Machine Learning, 2025

Abs arXiv HTML

Large Language Models (LLMs) with extended context windows are essential for complex tasks. We present LongRoPE2, a novel method that extends LLM context windows to 128K tokens while retaining 98.5% short-context accuracy. Our approach introduces improved position encoding strategies that enable near-lossless context extension.
Preprint

Hybrid-Gym: Training Coding Agents to Generalize Across Tasks

Yiqing Xie, Emmy Liu, Gaokai Zhang, Nachiket Kotalwar, Shubham Gandhi, Sathwik Acharya, and 4 more authors

arXiv preprint, 2026

Abs arXiv HTML Website

We present Hybrid-Gym, a framework for training coding agents to generalize across repository-level environments through synthetic task generation.