Siyu Yan 颜思宇

I completed my B.S. in Computer Science and Technology at Central South University (2022-2026), advised by Prof. Alex Jinpeng Wang. I will join The Hong Kong University of Science and Technology as an MPhil student in Fall 2026, supervised by Prof. Song Guo.

My research focuses on Agent Harness & RL and multimodal large models. I was a research assistant at Tsinghua University, THUNLP, advised by Prof. Maosong Sun and Prof. Zhiyuan Liu, and I am currently a research intern at Shanghai AI Laboratory, advised by Prof. Yihao Liu.

News

[07/31] Released IdeaFactory, an autonomous research ideation system with adaptive branching chain search.

[07/29] Released Your LLM, Your Style, a framework for measuring and controlling LLM behavioral modes.

[07/16] Released See2Think, a benchmark for diagnosing intermediate visual state use in multimodal models.

[06/26] Released DataEvolver, a self-evolving multi-agent data construction framework.

[01/25] One paper accepted by ACL 2025.

Publications

Agent Harness & RL

IdeaFactory adaptive research ideation workflow

IdeaFactory: Autonomous Research Ideation with Adaptive Branching Chain Search

Haiying Xu^*, Zixuan Wang^*, Siyu Yan^*, Haoze Liu, Shuo Cao, Siyu Zhang, Qinghua Xing, Chenhan Ji, Yihao Liu, Chengwei Qin

^* Equal contribution · AAAI 2027 Conference Submission

Paper

AgentRM: Enhancing Agent Generalization with Reward Modeling

Yu Xia, Jingru Fan, Weize Chen, Siyu Yan, Xin Cong, Zhong Zhang, Yaxi Lu, Yankai Lin, Zhiyuan Liu, Maosong Sun

ACL 2025

arXiv / Code

Multimodal Large Models

See2Think: Do Multimodal Models Really Use Intermediate Visual States?

Siyu Yan, Zhuoran Yan, Haiying Xu, Panhao Zhou, Jingyu Chen, Chenhao Ji, Shuo Cao, Yongheng Zhang, Haoze Liu, Siyu Zhang, Xiwen Gu, Yihao Liu, Alex Jinpeng Wang

Preprint 2026

Project Page / Paper / Code

DataEvolver: Self-Evolving Multi-Agent Data Construction for Text-Rich Image Generation

Siyu Yan, Yizhen Gao, Yilin Wang, Dongxing Mao, Alex Jinpeng Wang

arXiv 2026

Project Page / arXiv / Code

Others

Situated behavioral data and Behavioral Mode Axes framework

Your LLM, Your Style: Behavioral Mode Axes for LLM Behavioral Control

Haoze Liu, Run Liu, Haiying Xu, Jiahui Han, Siyuan Fang, Siyu Yan, Huiqi Deng, Guanchu Wang, Na Zou

AAAI 2027 Conference Submission

Paper

Experience

Tsinghua University, THUNLP Lab

Research Assistant, advised by Prof. Maosong Sun and Prof. Zhiyuan Liu

Jul. 2024 — Mar. 2026Beijing

Shanghai AI Laboratory

Research Scientist Intern, advised by Prof. Yihao Liu

Apr. 2026 — PresentShanghai

XbotPark Winter Camp

Team Leader

Winter 2024Shenzhen

AdventureX 2026

Creator

2026Hangzhou

Honors and Awards

Total of 11 national-level honors, all as first author or project lead. Selected awards:

National 1st PrizeNational College Innovation and Entrepreneurship Project; 1st in Central South University.
National 2nd PrizeChina Engineering Practice and Innovation Ability Competition; 1st in Hunan Province.
National 2nd PrizeNational Collegiate Computer Design Contest; 1st in Hunan Province.
National 2nd PrizeCross-Strait Computer Works Competition; 1st in Hunan Province.

Entrepreneurship

Engineering Training Virtual Simulation Platform

An interactive engineering training platform combining 3D reconstruction, physical simulation, and digital twins. Funded by the first batch of Hunan Province College Student Innovation and Entrepreneurship Fund.