Ph.D. Student East China Normal University, Shanghai, China Research Intern (Project UP 青云计划) Hunyuan (LLM Department), Tencent, Shanghai, China Email: yangfuli1215@gmail.com [Google Scholar]    [GitHub]

Research Interests

Visual Reasoning, Post-training for LLMs/MLLMs/UMMs.

Experience

• [2026-Now] I work as a research intern (Project UP) @Hunyuan (LLM Department), Tencent.

• [2025-2026] I worked as a research intern (Plan A) @Bailing (Ming Team), Ant Group, supervised by Dr. Ziyuan Huang.

• [2024-Now] I am pursuing my Ph.D. @LPAIS, ECNU, supervised by Prof. Yue Lyu, with research interests in MLLMs.

• [2023-2024] I worked as a research visitor @XMU, with research interests in Multi-modal information processing.

• [Previously] I received my B.E. from FZU.

News

• [2026-05] 🏅 I am selected as CVPR 2026 Golden Reviewer.

• [2026-05] 🎉 We release PFlowNet, a variational RL strategy for vision-centric reasoning.

• [2026-02] 🔥 We release DeepScan, a training-free framework for agentic visual reasoning.

• [2026-02] 🎇 Releases Ming-flash-Omni 2.0, an advanced UMM for unified visual understanding and generation.

• [2025-10] 🏅 I was recognized as an ICCV 2025 Outstanding Reviewer.

• [2025-09] 🎉 We release MoB, a vision token pruning method for efficient visual reasoning.

• [2025-09] 🏅 I am recognized as a NeurIPS 2025 Top Reviewer.

• [2025-07] 🎉 We release MSA², a mixture prototype learning paradigm for open-set text recognition.

• [2025-05] 🏅 I was recognized as an ICLR 2025 Notable Reviewer.

• [2024-09] 🎉 We release FaRE, a radical-based prototype learning paradigm for open-set character recognition.

• [2024-07] 🎉 We release Free Lunch, a frame-level contrastive learning method for low-quality text recognition.

• [2024-04] 🔥 We release DS-TDNN, a hybrid architecture with frequency-domain attention for speech processing.

Selected Projects

* Equal Contribution

♠ Multimodal Large Language Model (MLLM)

• Perceptual Flow Network for Visually Grounded Reasoning

  Yangfu Li*, Yu-ning Gong*, Teng Li, Yuanhuiyi Lyu, Tianyi Chen, Qi Liu, Ziyuan Huang, Zhihang Zhong, Dandan Zheng, Yue Lyu

  ICML 2026

  [Paper]

• StruVis: Enhancing Reasoning-based Text-to-Image Generation via Thinking with Structured Vision

  Yuanhuiyi Lyu, Kaiyu Lei, Ziqiao Weng, Xu Zheng, Lutao Jiang, Teng Li, Yangfu Li, Ziyuan Huang, Linfeng Zhang, Xuming Hu

  arXiv preprint 2026

  [Paper]

• TC-AE: Unlocking Token Capacity for Deep Compression Autoencoders

  Teng Li, Ziyuan Huang, Cong Chen, Yangfu Li, Yuanhuiyi Lyu, Dandan Zheng, Chunhua Shen, Jun Zhang

  arXiv preprint 2026

  [Paper]

• DO-Bench: An Attributable Benchmark for Diagnosing Object Hallucination in Vision-Language Models

  JiYang Wang, Jiawei Chen, Mengqi Xiao, Yu Cheng, Yangfu Li, Zhaoxia Yin

  arXiv preprint 2026

  [Paper]

• DeepScan: A Training-Free Framework for Visually Grounded Reasoning in Large Vision-Language Models

  Yangfu Li, Hongjian Zhan, Jiawei Chen, Yu-ning Gong, Qi Liu, Yue Lyu

  CVPR 2026

  [Paper] [Code 🌟230+]

• Why 1 + 1 < 1 in Visual Token Pruning: Beyond Naive Integration via Multi-Objective Balanced Covering

  Yangfu Li, Hongjian Zhan, Tianyi Chen, Qi Liu, Yu-jie Xiong, Yue Lyu

  NeurIPS 2025

  [Paper] [Code]

♠ Optical Character Recognition

• MSA²: Multi-task Framework with Structure-aware and Style-adaptive Character Representation for Open-set Chinese Text Recognition

  Yangfu Li, Hongjian Zhan, Qi Liu, Li Sun, Yu-jie Xiong, Yue Lyu

  ICCV 2025

  [Paper] [Code]

• FaRE: A Feature-Aware Radical Encoding Strategy for Zero-Shot Chinese Character Recognition

  Hongjian Zhan, Yangfu Li, Yu-jie Xiong, Yue Lyu

  ACCV (Oral) 2024

  [Paper]

• Free Lunch: Frame-level Contrastive Learning with Text Perceiver for Robust Scene Text Recognition in Lightweight Models

  Hongjian Zhan*, Yangfu Li*, Yu-jie Xiong, Umapada Pal, Yue Lyu

  ACM MM 2024

  [Paper] [Code]

♠ Speech Processing

• Dual-stream Time-Delay Neural Network with Dynamic Global Filter for Speaker Verification

  Yangfu Li, Jiapan Gan, Xiaodan Lin, Hongjian Zhan, Yingqiang Qiu, Hui Tian

  IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2023

  [Paper] [Code 🌟40+]

• UniVR: A Unified Framework for Pitch-Shifted Voice Restoration in Speaker Identification

  Yangfu Li, Xiaodan Lin

  APSIPA ASC 2023

  [Code]

♠ Image Processing

• DeflickerCycleGAN: Learning to Detect and Remove Flickers in a Single Image

  Xiaodan Lin*, Yangfu Li*, Jianqin Zhu, Huanqiang Zeng

  IEEE Transactions on Image Processing, 2023

  [Paper] [Code]

Services

Reviewer for ICML, NeurIPS, ICLR, CVPR, ECCV, ICCV, and AAAI.