Jian Hu

I am currently a Research Scientist in Tencent, I finish my PhD degree and Postdoc within the Computer Vision Group in the School of Electronic Engineering and Computer Science at Queen Mary, University of London, supervised by Prof. Shaogang Gong.

Before joining the Computer Vision Group in QMUL, I was in Shanghai Jiao Tong University, supervised by Prof. Hongya Tuo and working closely with Prof. Junchi Yan.

Currently, my research focuses on deep learning and computer vision, with particular emphasis on test-time training and adaptation of Multi-modal Large Language Models (MLLMs) and their applications.

Email  /  Google Scholar  /  Website  /  Github

Research

I am interested in machine learning and computer vision. My current focus is on improving MLLMs' performance in real-world scenarios with limited supervision through test-time training and adaptation.

News

• 2026-08: Our survey on multi-modal coding is accepted by TMLR.
• 2025-12: I serve as the Proceedings Chair on the Organising Committee of BMVC 2026.
• 2025-11: The BMVA Multimodal Large Models workshop that I co-chaired was successfully held on 5 November 2025.
• 2025-09: URPA is accepted to NeurIPS 2025.
• 2025-04: INT is accepted to IJCAI 2025.
• 2025-04: One paper is accepted to TMM.
• 2025-04: Give a talk in OpenCompass on our V-STaR benchmark.
• 2025-03: Released a curated multimodal reasoning MLLMs collection repository.
• 2024-11: Be awarded as a top reviewer in NeurIPS 2024.
• 2024-09: ProMaC is accepted to NeurIPS 2024.
• 2024-06: Start a research internship in Spotify.
• 2024-04: Give a talk in BMVA workshop: Trustworthy Multimodal Learning with Foundation Models: Bridging the Gap between AI Research and Real World Applications.
• 2023-12: GenSAM is accepted to AAAI 2024.
• 2023-05: UHPKD is accepted to SIGIR 2023.

Preprint

	ViSMaP: Unsupervised Hour-long Video Summarisation by Meta-Prompting Jian Hu, Dimitrios Korkinof, Shaogang Gong, Mariano Beguerisse-Diaz Under review arXiv A system to summarise hour long videos with no supervision.
	CoS: Chain-of-Shot Prompting for Long Video Understanding Jian Hu, Zixu Cheng, Chenyang Si, Wei Li, Shaogang Gong Under review arXiv / website / code To frame shot selection as test-time visual prompt optimisation.

Publications

	BeyondNL2Code: A Structured Survey of Multimodal Code Intelligence Xuanle Zhao, Qiushi Sun, Jingyu Xiao, Xuexin Liu, Haoyue Yang, Qiaosheng Chen, Xianzhen Luo, Jing Huang, Yufeng Zhong, Lei Chen, Shuai Fu, Zhenlin Wei, Jinhe Bi, Lei Jiang, Haibo Qiu, Siqi Yang, Peng Shi, Jian Hu (Corresponding), Zhixiong Zeng (Corresponding) TMLR, 2026 arXiv A structured survey of multimodal code intelligence beyond NL2Code.
	V-STaR: Benchmarking Video-LLMs on Video Spatio-Temporal Reasoning Zixu Cheng, Jian Hu (Corresponding), Ziquan Liu, Chenyang Si, Wei Li, Shaogang Gong CVPRW, 2026 arXiv / website / code To decompose video understanding into a Reverse Spatio-Temporal Reasoning (RSTR) task that simultaneously evaluates what objects are present, when events occur, and where they are located while capturing the underlying Chain-of-Thought (CoT) logic.
	Uncertainty-quantified Rollout Policy Adaptation for Unlabelled Cross-domain Temporal Grounding Jian Hu, Zixu Cheng, Shaogang Gong, Isabel Guan, Jianye Hao, Jun Wang, Kun Shao NeurIPS, 2025 arXiv An unsupervised RL-based cross-domain temporal grounding approach.
	INT: Instance-Specific Negative Mining for Task-Generic Promptable Segmentation Jian Hu, Zixu Cheng, Shaogang Gong IJCAI, 2025 (oral) arXiv To adaptively reduce the influence of irrelevant (negative) prior knowledge whilst increasing the use of the most plausible prior knowledge, selected by negative mining with higher contrast, in order to optimise instance-specific prompt generation.
	Class-Aware Diversified Augmentation for Open-Set Single Domain Generalization Jian Hu, Shaogang Gong, Weitong Cai, Junchi Yan IEEE Transactions on Multimedia, 2025 Paper Explore single domain open-set generalization with class-aware augmentation.
	Leveraging Hallucinations to Reduce Manual Prompt Dependency in Promptable Segmentation Jian Hu, Jiayi Lin, Junchi Yan, Shaogang Gong NeurIPS, 2024 arXiv / website / code Using hallucinations as prior knowledge to help create specific prompts for segmenting tasks, reducing the need for manual prompts.
	Relax Image-Specific Prompt Requirement in SAM: A Single Generic Prompt for Segmenting Camouflaged Objects Jian Hu*, Jiayi Lin*, Weitong Cai, Shaogang Gong AAAI, 2024 arXiv / website / code Eliminate the need for manual prompts for SAM in various challenging segmentation tasks.
	Uncertainty-based Heterogeneous Privileged Knowledge Distillation for Recommendation System Ang Li*, Jian Hu*, Ke Ding, Xiaolu Zhang, Jun Zhou, Yong He SIGIR, 2023 Paper Proposing a novel algorithm to address heterogeneous knowledge distillation-based transfer learning in industrial recommendation systems.
	Global-Aware Model-Free Self-distillation for Recommendation System Ang Li*, Jian Hu*, Lu Wei, Ke Ding, Xiaolu Zhang, Jun Zhou, Yong He DASFAA, 2023 Paper Introducing a novel algorithm called Global-aware Model-free Self-Distillation to address label noise in training data in the Alipay advertising system.
	Learning Unbiased Transferability for Domain Adaptation by Uncertainty Modeling Jian Hu*, Haowen Zhong*, Fei Yang, Shaogang Gong, Guile Wu, Junchi Yan ECCV, 2022 arXiv / code Delving into the transferability estimation problem in domain adaptation and proposing a non-intrusive Unbiased Transferability Estimation Plug-in (UTEP) by modeling the uncertainty of a discriminator in adversarial-based DA methods to optimize unbiased transfer.
	Attribute-Conditioned Face Swapping Network for Low-Resolution Images Ang Li*, Jian Hu*, Chilin Fu, Xiaolu Zhang, Jun Zhou ICASSP, 2022 Paper A novel Attribute-Conditioned Face Swapping Network (AFSNet) to preserve attributes and handle low resolution images.
	Domain Adaptive YOLO for One-Stage Cross-Domain Detection Shizhao Zhang, Hongya Tuo, Zhongliang Jing, Jian Hu ACML, 2021 arXiv Improving cross-domain performance for one-stage detectors: image level feature alignment is used to strictly match local features and loosely match global features.
	Discriminative Partial Domain Adversarial Network Jian Hu, Hongya Tuo, Chao Wang, Lingfeng Qiao, Haowen Zhong, Junchi Yan, Zhongliang Jing, Henry Leung ECCV, 2020 Paper Addressing the partial domain adaptation problem with a discriminative partial domain adversarial network with theoretical analysis.
	Unsupervised Satellite Image Classification based on Partial Transfer Learning Jian Hu, Hongya Tuo, Chao Wang, Haowen Zhong, Pan Han, Lingfeng Qiao, Zhongliang Jing Aerospace Systems, 2019 Paper Focusing on how to achieve high accuracy on unsupervised satellite image classification.
	Multi-Weight Partial Domain Adaptation Jian Hu, Hongya Tuo, Chao Wang, Lingfeng Qiao, Haowen Zhong, Zhongliang Jing BMVC, 2019 (spotlight) Paper Focusing on how to transfer knowledge from a massive labelled dataset to an unlabelled miniature one.

Service

Special thank you to the source code.