I am in the second year of my PhD at the Shanghai Innovation Institute, supervised by Dr. Qipeng Guo at the Shanghai AI Laboratory and Prof. Xipeng Qiu at FudanNLP, and supervised by Prof. Chenhui Li at the Shanghai Institute of AI Education, East China Normal University. I spent my first PhD year with the InternLM team at the Shanghai Artificial Intelligence Laboratory. I completed my Academic Master’s degree in Computer Science and Technology at East China Normal University supervised by Prof. Chenhui Li and Prof. Changbo Wang (2021-2024) and my undergraduate studies in Mathematics & Applied Mathematics and Computer Science at East China University of Science and Technology (2017-2021). My research focuses on cross-modal LLM Agents, multimodal image editing, and computer graphics. I have served as a reviewer for top-tier conferences including IEEE VIS, CVPR and NeurIPS.

🔥 News

• 2025.07:  🎉 1 paper accepted to ACM MM 2025.
• 2025.07:  🎉 1 paper accepted to IEEE VIS 2025.
• 2025.05:  🎉 1 paper accepted to ICML 2025.
• 2025.05:  🎉 1 paper accepted to ACL (main) 2025.
• 2024.03:  🎓 Secured a admission offer through the Spring Camp selection at the Shanghai Innovation Institute; expected to enroll in September 2025.
• 2024.09:  🎓 Starting my PhD at Shanghai Institute of AI Education, East China Normal University.
• 2024.05:  💼 Joined the InternLM group at Shanghai Artificial Intelligence Laboratory as a LLM Research Intern.

🔧 Academic Services

📝 Publications

ICML 2025

TextCenGen: Attention-Guided Text-Centric Background Adaptation for Text-to-Image Generation
Tianyi Liang^†, Jiangqi Liu^†, Yifei Huang, Shiqi Jiang, Jianshen Shi, Changbo Wang, and Chenhui Li*.
ICML 2025, Accepted (CCF A)
arXiv  |  Code  |  Project Page

IEEE VIS 2025

Sel3DCraft: Interactive Visual Prompts for User-Friendly Text-to-3D Generation
Nan Xiang^†, Tianyi Liang^†, Haiwen Huang, Shiqi Jiang, Hao Huang, Liangyu Chen, Changbo Wang, and Chenhui Li*
IEEE Transactions on Visualization and Computer Graphics, 2026 (IEEE VIS’25) (CCF A, JCR Q1)
arXiv

ACL 2025

CritiQ: Mining Data Quality Criteria from Human Preferences
Honglin Guo, Kai Lv, Qipeng Guo, Tianyi Liang, Zhiheng Xi, Demin Song, Qiuyinzhe Zhang, Yu Sun, Kai Chen, Xipeng Qiu, Tao Gui*
ACL 2025, Accepted (Main Conference)
arXiv  |  Code

MM 2025

Music2Palette: Emotion-aligned Color Palette Generation via Cross-Modal Representation Learning
Jiayun Hu, Yueyi He, Tianyi Liang, Changbo Wang, Chenhui Li*
MM 2025, Accepted (CCF A) arXiv

arXiv

IFDECORATOR: Wrapping Instruction Following Reinforcement Learning with Verifiable Rewards
Xu Guo^†, Tianyi Liang^†, Tong Jian, Xiaogui Yang, Ling-I Wu, Chenhui Li, Zhihui Lu, Qipeng Guo, Kai Chen*
^†Equal contribution
arXiv  |  Code  |  Project Page

arXiv

Consensus Entropy: Harnessing Multi-VLM Agreement for Self-Verifying and Self-Improving OCR
Yulong Zhang^†, Tianyi Liang^†, Xinyue Huang, Erfei Cui, Xu Guo, Pei Chu, Chenhui Li, Ru Zhang, Gongshen Liu*
^†Equal contribution
arXiv  |  Code

arXiv

GAOKAO-Eval: Does high scores truly reflect strong capabilities in LLMs?
Zhikai Lei^†, Tianyi Liang^†, Hanglei Hu^†, Jin Zhang, Yunhua Zhou, Yunfan Shao, Linyang Li, Chenchui Li, Changbo Wang, Hang Yan*, Qipeng Guo
^†Equal contribution
arXiv  |  Code  |  Model

💼 Internships

• 2024.5 - Present: Shanghai Artificial Intelligence Laboratory, LLM Algorithm Research Intern
  • Research on Open-Compass GAOKAO evaluation and InternLM2-WQX-VL-20B
  • Developed an innovative Verifier-Guided workflow for quality control of Web Data and PDF corpora in InternLM3 project. Proposed a novel reflection-based criteria optimization method, improving bad case detection F1-score from 66% to 86%. More details can be found in CritiQ.
  • Conducting research on LLM-as-Judge and synthetic data generation for large language model pre-training. Implemented domain filtering and Verifier-Guided Rephrase techniques, reducing required training data by 85% while maintaining model performance.

💻 Projects

Open Source Course Sharing Initiatives

• ECUST-CourseShare

  
  A comprehensive course resource sharing platform for East China University of Science and Technology. This project aims to build a collaborative platform where students can share course materials, exam papers, and study resources. It covers various departments and provides historical exam papers, review materials, teacher courseware, and homework answers.

• ECNU-PGCourseShare

  
  Graduate course resource sharing platform for East China Normal University. This project extends beyond basic course materials to include academic intelligence resources, helping graduate students navigate research fields, discover collaboration opportunities, and access cutting-edge academic tools and best practices.

Community Platform

• ECNU Forum

  
  A modern forum platform based on Flarum for East China Normal University students. The live forum is accessible at ecnu.im, providing a space where ECNU students can freely communicate and exchange ideas. This is an unofficial community-driven platform that fosters academic and social interaction among students.

Research Projects

• GAOKAO-Eval

  
  An evaluation benchmark for large language models on Chinese GAOKAO tasks, providing datasets, evaluation pipeline, and reproducible scripts within the OpenCompass ecosystem.

🎖 Honors and Awards

• 2021.7 Shanghai Outstanding Graduate & Outstanding Thesis Award
• 2019.12 National College Student Mathematical Modeling Competition, Shanghai Region First Prize
• 2019.12 National Scholarship
• 2018.12 Shanghai Scholarship