I am Anpeng Wang (王安鹏), currently a Master’s student at the School of Control Science and Engineering, Shandong University. I am fortunate to study under the supervision of Professor Runmin Cong and am affiliated with the Multimedia and Vision Processing Group (MVP Group), which is attached to the Key Laboratory of Machine Intelligence and System Control, Ministry of Education.
I am interested in one of the most basic and distinctive capabilities of human intelligence: the ability to adapt to unfamiliar environments, rapidly and flexibly. Humans can generalize from prior experience, learn rapidly from a few new examples, and use these abilities to acquire surprisingly rich knowledge about the world.
I study how models can develop similar capabilities. How can models generalize out of distribution? How can they effectively reason with available evidence and prior knowledge? When existing knowledge is insufficient, how can they rapidly learn new concepts and capabilities from a few examples? And how should we train models to develop these abilities?
My current research approaches these questions through vision and multimodal models, with a particular focus on few-shot learning, cross-domain learning, and visual reasoning. I am also broadly interested in in-context learning, continual learning, and knowledge internalization.
Feel free to contact me at rawwap@mail.sdu.edu.cn for academic discussions or collaborations.
🔥 News
- 2026.03: 🎉🎉🎉 A paper has been accepted by IEEE ICME 2026!
- 2026.02: 🎉🎉🎉 A cooperative paper has been accepted by CVPR 2026!
- 2026.01: Attend AAAI 2026 in Singapore and give an oral presentation.
- 2025.11: 🎉🎉🎉 My first paper has been accepted by AAAI 2026 (Oral)!
📝 Publications
Divide-and-Conquer Decoupled Network for Cross-Domain Few-Shot Segmentation
Runmin Cong, Anpeng Wang, Bin Wan, Cong Zhang, Xiaofei Zhou, Wei Zhang
- This paper focuses on Cross-Domain Few-Shot Segmentation, leveraging feature disentanglement enabled by contrastive-adversarial learning.
ADSeeker: A Knowledge-Grounded Reasoning Framework for Industry Anomaly Detection and Reasoning
Kai Zhang, Zekai Zhang, Xihe Sun, Anpeng Wang, et al.
- This paper proposes the ADSeeker framework to improve fine-grained reasoning and zero-shot performance for industrial anomaly detection.
- We construct the first visual document knowledge base SEEK-M&V and the largest anomaly detection dataset MulA to address data scarcity and insufficient type-level annotations.
P³-SAM: SAM with Perceptual Parallel Prompt for Few-Shot Strip Steel Surface Defect Segmentation
Qian Xu, Hang Xiong, Anpeng Wang*, et al. ( * corresponding author )
- This paper proposes the P³-SAM model, which optimizes SAM via two core strategies of POE and PPG to address the challenges of few-shot strip steel surface defect segmentation.
🎖 Honors and Awards
- 2026.10 Top Reviewer of NeurIPS 2026
- 2026.03 Shandong University Outstanding Graduate
- 2026.01 Shandong University OpenAtom Open Source Pioneer Scholarship
- 2024.09 Shandong University Outstanding Student Scholarship
- 2024.07 Shandong University Innovation and Entrepreneurship Scholarship
- 2023.12 APMCM Asia-Pacific Mathematical Contest in Modeling First Prize
- 2023.09 Shandong University Outstanding Student Scholarship
- 2023.08 National University Students Intellingent Car Race National Second Prize
📖 Educations
- 2026.09 - present, Master, School of Control Science and Engineering, Shandong University, Jinan, China. Advisor: Prof. Runmin Cong
- 2022.09 - 2026.07, Undergraduate, School of Control Science and Engineering, Shandong University, Jinan, China. Advisor: Prof. Runmin Cong
💻 Professional Services
Reviewer | PC Member
- Conference on Neural Information Processing Systems (NeurIPS) 2026
- Association for the Advancement of Artificial Intelligence (AAAI) 2026-2027
- IEEE International Conference on Multimedia and Expo (ICME) 2026