Chengchun Liu 刘丞军 · 劉丞軍

AI for chemistry. Built to reason.

Toward scientific AI that connects measurement, mechanism, and discovery.

Explore selected work

Google Scholar GitHub ORCID LinkedIn Instagram

Research focus

Scientific agents Molecular spectra Geometric learning Autonomous chemistry

01 / Selected work

Selected publications

Research across molecular structure elucidation, quantum geometry, interpretable learning, and chemical automation.

Angewandte Chemie International Edition · 2026

Hypothesis-and-Refinement Learning of Organic Structures From Multimodal Spectroscopic Data

Chengchun Liu^†, Zhiyuan Yan^†, Li Yuan, Hao Li, Boxuan Zhao, Yonghong Tian, Bartosz A. Grzybowski, and Fanyang Mo

A hypothesis-and-refinement system that proposes molecular structures from multimodal spectra, then repairs them with formula-, exact-mass-, and unsaturation-aware molecular generation.

View paper

01 Hypothesis

Spectra → hypothesis → refinement

Journal of Chemical Theory and Computation · 2026

A Cross-Domain Graph Learning Protocol for Single-Step Molecular Geometry Refinement

Chengchun Liu, Wendi Cai, Boxuan Zhao, and Fanyang Mo

GeoOpt-Net uses SE(3)-equivariant, multifidelity learning to map inexpensive force-field conformers directly toward DFT-quality molecular geometries in one deterministic forward pass.

View paper View code

Force-field conformer DFT-quality geometry

SE(3)

One deterministic refinement step

Multi-variable system

x₁x₂x₃x₄x₅xₙ

→

Hierarchical discovery

x₁x₂x₃x₄

↓

x₁x₂sin x₃

↓

Interpretable law ŷ = αΦ₁ + βΦ₂

compact · explicit · interpretable

Many variables → hierarchical discovery → compact equation

Advanced Science · 2026

Unsupervised Hierarchical Symbolic Regression for Interpretable Property Modeling in Complex Multi-Variable Systems

Siyu Lou^†, Chengchun Liu^†, Dongxiao Zhang, Yuntian Chen, and Fanyang Mo

A layered symbolic-learning framework that turns complex multivariable relationships into compact, chemically interpretable equations.

View paper

Molecule10⁻¹⁰ m

Surface10⁻⁶ m

Column10⁻¹ m

Closed-loopAI + automation

From molecular interactions to closed-loop separation

Accounts of Chemical Research · 2026

Automation and AI-Powered Prediction in Chromatographic Separation

Chengchun Liu and Fanyang Mo

A research perspective connecting automated experimentation, retention prediction, condition optimization, and closed-loop chromatographic separation.

View paper

Structural neighborhood

Analytical Chemistry · 2024

Infrared Spectra Prediction for Functional Group Region Utilizing a Machine Learning Approach With Structural Neighboring Mechanism

Chengchun Liu, Ruqiang Zou, and Fanyang Mo

A functional-group-centered learning strategy that uses local structural context to improve the accuracy, robustness, and interpretability of infrared-spectrum prediction.

View paper View code

^† These authors contributed equally.

Looking for the complete publication record, including collaborative work?

Open Google Scholar

02 / About

Chemistry, computation, and systems that can close the loop.

I am a Ph.D. candidate in Materials Physics and Chemistry at Peking University, advised by Prof. Fanyang Mo . Before that, I studied chemistry at Nankai University.

My research asks how AI can work with the evidence chemists actually use: spectra, three-dimensional structures, physical constraints, experimental conditions, and iterative feedback.

Aug 2026 — Present

ByteDance Seed

Seed Edge intern · STEM Scientist Program

Sep 2022 — Present

Peking University

Ph.D. candidate in Materials Physics and Chemistry · Expected graduation: Jun 2027

Sep 2018 — Jun 2022

Nankai University

B.Sc. in Chemistry

03 / Contact

Let’s talk about scientific AI.

I’m interested in collaborations around molecular reasoning, chemistry agents, spectroscopy, geometric learning, and research automation.

liuchengchun@stu.pku.edu.cn