Jifeng Song's Homepage

About Me

Hi, I am Jifeng Song, a fourth-year Ph.D. student in Electrical and Computer Engineering at the University of Pittsburgh and also a research assistant in the Cancer Virology Program at UPMC Hillman Cancer Center, co-advised by Prof. Yufei Huang and Prof. Zhi-Hong Mao. I received my B.E. in Electrical Engineering and Automation from Huazhong University of Science and Technology.

My current research focuses on post-training and efficient inference for multimodal large language models (MLLMs), with applications in scientific discovery and cancer research. I am particularly interested in scientific document understanding and multimodal knowledge discovery.

I am actively seeking internship opportunities starting in 2027. You can find my CV here.

News

[08/2026] 🎉 Paper accepted to EMNLP Main: FigEx2: Visual-Conditioned Panel Detection and Captioning for Scientific Compound Figures.
[05/2026] 🎉 Paper accepted to MLSys: Attribution-based Sparse Activation in Large Language Models.
[04/2026] New preprint: spatiAlytica.
[01/2026] New preprint: Aligning Findings with Diagnosis.
[12/2025] New preprint: A Process-Centric Survey of AI for Scientific Discovery.
[08/2025] 🎉 Paper accepted to EMNLP Findings: FigEx: Aligned Extraction of Scientific Figures and Captions.
[06/2024] New preprint: Achieving Sparse Activation in Small Language Models.

Selected Publications | Full list

EMNLP 2026

FigEx2: Visual-Conditioned Panel Detection and Captioning for Scientific Compound Figures

Jifeng Song, Arun Das, Pan Wang, Hui Ji, Kun Zhao, Yufei Huang

Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing

Paper Code

MLSys 2026

Attribution-based Sparse Activation in Large Language Models

Jifeng Song^†, Xiangyu Yin^†, Boyuan Yang, Kai Huang, Weichen Liu, Wei Gao (^†Equal contribution)

Proceedings of Machine Learning and Systems (MLSys) 2026

Paper Code

EMNLP 2025

FigEx: Aligned Extraction of Scientific Figures and Captions

Jifeng Song, Arun Das, Ge Cui, Yufei Huang

Findings of the Association for Computational Linguistics: EMNLP 2025

Paper Code Dataset

bioRxiv

spatiAlytica: Viewer-Grounded Multimodal Agentic System for Interactive Spatial Omics Analysis

Arun Das, Kexun Zhang, Jifeng Song, Meiru Han, Angela Chen, Wen Meng, Hugh Galloway, Po-Yuan Chen, Sumin Jo, Zhentao Liu, Md Musaddaqul Hasib, Adam Officer, Harsh Sinha, Yu-Chiao Chiu, Shou-Jiang Gao, Lei Li, Yufei Huang

bioRxiv preprint 2026

Paper

arXiv

Aligning Findings with Diagnosis: A Self-Consistent Reinforcement Learning Framework for Trustworthy Radiology Reporting

Kun Zhao, Siyuan Dai, Pan Wang, Jifeng Song, Hui Ji, Chenghua Lin, Liang Zhan, Haoteng Tang

arXiv preprint 2026

Paper