About

I am a first-year undergraduate student at Xi'an Jiaotong-Liverpool University, majoring in Applied Mathematics, and I am expected to complete my B.Sc. degree at the University of Liverpool in 2029. I am currently seeking research internships or visiting student opportunities. If you are interested in my research, please feel free to contact me.

My research interests span machine learning and mathematical reasoning. In my spare time, I also engage in research on multimodal large language models and geometric problem solving.

News

  • 2026.05I was supported by the Summer Undergraduate Research Fellowship (SURF) at XJTLU.

Education

B.S. in Applied Mathematics University of Liverpool. GPA: 4.0 / 4.0.
2025.09 — Present

Publications

* denotes equal contribution.

OmniGeo thumbnail

Omni-Geo: Full-Domain Geometry with Multimodal Diagram Generation

Ruoran Xu*, Wending Gao*, Haoyu Cheng*, Chengrui Zhang, Maizhen Ning, Qiufeng Wang

Under review

We present Omni-Geo, the first unified benchmark for general geometric intelligence spanning plane, analytic, and solid geometry, built on a standardized Geometric Description Language (GDL) and an SDF-based diagram synthesis engine that produces contamination-free data. Comprising ~23K problems evaluated on 11 state-of-the-art LLMs and MLLMs, Omni-Geo highlights the necessity of a comprehensive, unified benchmark for assessing geometric reasoning.

PhysElite thumbnail

PhysElite: How Far Are LLMs from Solving Olympiad-Level Physics Problems?

Ruoran Xu*, Wending Gao*, Liyunfeng Cheng*, Aixin Shi*, Haoyu Cheng*, Zixiang Fang, Yiqiang Zou, Qiufeng Wang

Under review

We present PhysElite, A large Olympic physics competition multimodal benchmark. We tracked the contestants' daily practice questions, modeled the multimodal problem-solving steps required to answer them, and analyzed current systems against expert-level performance.

FormalAnalyticGeo thumbnail

FormalAnalyticGeo: A Neural-Symbolic Based Framework for Multimodal Analytic Geometry Problem Generation

Ruoran Xu*, Wending Gao*, Qiufeng Wang

Under review

We present FormalAnalyticGeo, a scalable framework for fully automatic multimodal analytic geometry problem generation, built around a formal intermediate representation (CDL) that bridges free-form problem text with precise SDF-based diagram rendering. Four specialized LLM components — Generator, Formalizer, Measurer, and Quality Verifier — form a closed loop requiring no human annotation, yielding AnalyticGeo7K, a dataset of 7K+ verified problems with a median ground-truth relative error of 0.70%.

Projects

ApolloGeo: A Formalized Framework for Analytic Geometry Problem Solving

A formal framework for analytic geometry with CDL predicates and a conic theorem bank. Built AGPS, a solver unifying geometric relations and algebraic computation, and curated a large-scale formally-annotated dataset — substantially outperforming SOTA MLLMs and approaching human expert performance.

Expected submission: EMNLP 2026

Activities

Mathematical Contest in Modeling (MCM) Honorable Mention.
2026
Lecturer — Academic Department, Mathematics Club Organized and delivered multiple large-scale lectures on mathematics topics.
 

Skills

ProgrammingPython, C/C++, MATLAB, Lean
FrameworksPyTorch, TensorFlow
ResearchMathematical Reasoning, Autoformalization, Post Training
LanguagesChinese (native), English (fluent)

Contact

I am actively looking for research internships and visiting-student opportunities.

LocationSuzhou, China