Junming Liu

I am currently a Research Intern at Shanghai AI Lab, where I work on topics related to Generative Intelligence, Multimodal Reasoning, and Graph Theory. Prior to this, I received my Master’s degree in Computer Science from Tongji University, and my Bachelor’s degree in Intelligent Science from Dalian Maritime University.

My work seeks to enhance the creative fidelity and cognitive depth of AI systems, while ensuring logical consistency through knowledge representations. Recently, my research has been centered around the following areas:

Generative Modeling for Scientific Discovery.
Memory-Augmented Agents.
Post-training of Multimodal Large Language Models for Spatial Cognition.

I am actively seeking Research Assistant positions, Research Internships, and Collaborations. I am available for both on-site and remote opportunities. Please feel free to contact me.

News

May, 2026	Our paper Neuroevolution for Physical Dynamics (Evo-ManiEarth) has been accepted by IEEE Transactions on Evolutionary Computation (TEVC)! Congrats to all collaborators! 🎉🎉
Apr, 2026	Our survey “A Comprehensive Survey of Interaction Techniques in 3D Scene Generation” has been accepted by IJCAI 2026! We propose a unified taxonomy covering interactive generation, interactive editing, and embodied interaction. A curated list of related papers is available at Link. 🤗🤗
Jan, 2026	Our paper Domain-Adaptive Model Merging (DMM) has been accepted by ICASSP 2026! 🎉🎉
Jan, 2026	Our paper Adversarial Mutual Information Distillation (AMID) has been accepted by WWW 2026! 🎉🎉
Nov, 2025	Our paper ReBrain has been accepted by WACV 2026! 🎉🎉
Aug, 2025	Our paper Commonality-Oriented Gradient Optimization (COGO) has been accepted by PRCV 2025! 🎉🎉
Jul, 2025	Our paper Hierarchical Multi-Agent Retrieval-Augmented Generation (HM-RAG) has been accepted by ACM MM 2025! HM-RAG orchestrates a three-tier agent hierarchy to split complex queries and unify diverse modalities, advancing multimodal RAG reasoning. Code available at Link. 🤗🤗
Jun, 2025	Our paper Vision-align-to-Language integrated Knowledge Graph (VaLiK) has been accepted by ICCV 2025! VaLiK grounds vision in text and filters noise for annotation‑free MMKGs, boosting LLM reasoning to SOTA with high efficiency. Code available at Link. 🤗🤗
Jan, 2025	Join Shanghai AI Lab as a Research Intern, targeting Knowledge Reasoning!⚡️️⚡️

Selected Publications

WWW

AMID: Model-Agnostic Dataset Distillation by Adversarial Mutual Information Minimization

Aoqi Wu, Junming Liu, Yuwei Zhang, Weiquan Huang, Liang Hu^†, Yifan Yang, Qi Zhang, Jiaxing Miao, Yuhan Tang, and Zhongyuan Lai

Proceedings of the ACM on Web Conference, 2026

PDF
ACM MM

HM-RAG: Hierarchical Multi-Agent Multimodal Retrieval Augmented Generation

Pei Liu, Xin Liu, Ruoyu Yao, Junming Liu, Siyuan Meng, Ding Wang^†, and Jun Ma^†

Proceedings of the 33rd ACM International Conference on Multimedia, 2025

PDF Code
ICCV

Aligning Vision to Language: Annotation-Free Multimodal Knowledge Graph Construction for Enhanced LLMs Reasoning

Junming Liu, Siyuan Meng, Yanting Gao, Song Mao, Pinlong Cai, Guohang Yan, Yirong Chen, Zilin Bian, Ding Wang^†, and Botian Shi

Proceedings of the IEEE/CVF International Conference on Computer Vision, 2025

PDF Code