Large Language Models · Agents · Reasoning
Kou Shi
My work focuses on LLM agents, tool use, lifelong skill learning, and scientific agents.
Education
Education
University of Science and Technology of China USTC
M.S. Student
Hangzhou Dianzi University HDU
Bachelor's Degree
Research
Publications & Technical Reports
-
2026
FACET: Preserving Source Intent and Executable State in Terminal Task Synthesis
A framework for synthesizing high-quality terminal-agent training tasks by preserving source intent and grounding artifacts in a shared executable environment state.
arXiv:2608.18580 -
2026
AsyncTool: Evaluating the Asynchronous Function Calling Capability under Multi-Task Scenarios
A benchmark for asynchronous function calling, dependency tracking, and coordination under multi-task workloads and tool latency.
EMNLP 2026
arXiv:2605.27995 -
2026
Intern-S2-Preview: Scientific Agentic Foundation Model
A technical report on scientific agentic foundation models for multimodal understanding, reasoning, generation, and long-horizon tasks.
arXiv:2608.13505
Other Publications
-
arXiv
SkillFlow: Benchmarking Lifelong Skill Discovery and Evolution for Autonomous Agents
Ziao Zhang, Kou Shi, et al.
-
arXiv
SaaSBench: Exploring the Boundaries of Coding Agents in Long-Horizon Enterprise SaaS Engineering
Qingnan Ren, Shun Zou, Shiting Huang, Ziao Zhang, Kou Shi, et al.
-
arXiv
Internalizing Meta-Experience into Memory for Guided Reinforcement Learning in Large Language Models
Shiting Huang, Zecheng Li, Yu Zeng, Qingnan Ren, Zhen Fang, Qisheng Su, Kou Shi, et al. · EMNLP 2026
Internships
Internships
vivo Hangzhou R&D Center
Algorithm Intern, contributing to foundation model iteration
AI Lab
Foundation Model Intern, contributing to Intern-S2-Preview model iteration
Collaborative Research
Collaborative Research
Research involving my broader group is listed separately from papers on which I am a named author.