Ph.D. Student • East China Normal University
Yangfu Li
School of Communication & Electronic Engineering, East China Normal University
I am Yangfu Li, a Ph.D. student in Information and Communication Engineering at East China Normal University. I am advised by Prof. Yue Lyu and Prof. Zhi-Qin John Xu (@SJTU). Previously, I obtained my B.Eng. in Electronic Engineering from Fuzhou University. Currently, I focus on multi-modal large language models (MLLMs), especially visually grounded reasoning and visual token pruning, as well as OCR for open-set text recognition, low-quality text recognition, and handwritten spotting.
About Me
My research centers on multimodal intelligence, with a current emphasis on visually grounded reasoning in MLLMs and efficient visual token processing. I am also interested in OCR, especially open-set text recognition, low-quality text recognition, and handwritten spotting.
Multi-modal Large Language Models (MLLMs)
Visually grounded reasoning, visual token pruning, and multimodal perception-reasoning alignment.
Optical Character Recognition (OCR)
Open-set text recognition, low-quality text recognition, and handwritten spotting.
Education
Internship & Visitor Experience
Publications
-
MLLMs
Perceptual Flow Network for Visually Grounded ReasoningSubmitted to ICML (Under Review), 2026 -
MLLMs
DeepScan: A Training-Free Framework for Visually Grounded Reasoning in Large Vision-Language ModelsThe IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026 -
MLLMs
Why 1 + 1 < 1 in Visual Token Pruning: Beyond Naive Integration via Multi-Objective Balanced CoveringThe Thirty-Ninth Annual Conference on Neural Information Processing Systems (NeurIPS), 2025 -
OCR
MSA²: Multi-task Framework with Structure-aware and Style-adaptive Character Representation for Open-set Chinese Text RecognitionInternational Conference on Computer Vision (ICCV), 2025 -
OCR
FaRE: A Feature-Aware Radical Encoding Strategy for Zero-Shot Chinese Character RecognitionAsian Conference on Computer Vision (ACCV Oral), 2024 -
OCR
Free Lunch: Frame-level Contrastive Learning with Text Perceiver for Robust Scene Text Recognition in Lightweight ModelsACM International Conference on Multimedia (ACM MM), 2024 -
Speech Processing
Dual-stream Time-Delay Neural Network with Dynamic Global Filter for Speaker VerificationIEEE/ACM Transactions on Audio, Speech, and Language Processing (TASLP), 2023 -
Speech Processing
UniVR: A Unified Framework for Pitch-Shifted Voice Restoration in Speaker IdentificationAsia Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), 2023 -
Image Restoration
DeflickerCycleGAN: Learning to Detect and Remove Flickers in a Single ImageIEEE Transactions on Image Processing (TIP), 2023