Hi, There👋! I am a student at Huazhong University of Science and Technology(HUST).

I am currently diving into Few-shot Learning and Visual Language Model, aiming to help build more intelligent systems.

Feel free to reach me via email or WeChat: ABC-4941

🔥 News

2026.03: 🎉🎉 We win the 2nd Place in the NTIRE 2026 CD-FSOD Challenge, CVPR 2026.
2026.02: 🎉🎉 One paper “Interpretable CDSFSL with Rectified Target-Domain Local Alignment”, was accepted by CVPR 2026.

📝 Publications

Under Review

Yaze Zhao^†, Yicong Liu^†, Yixiong Zou, Yuhua Li, Ruixuan Li

By establishing fine-tuning baselines of CLIP for CDFSL, we find adapter-based methods consistently outperform prompt-based ones—contrary to in-domain scenarios. We analyze this phenomenon and discover LoRA’s superiority stems from rectifying the collapsed attention of visual [CLS] token, enhancing modality alignment and class separation by focusing on text-related visual regions. Further, we find textual [EOS] token exhibit much better attention to visual samples, and CLIP’s standard contrastive loss weakly constrains modality alignment.
Based on these insights, we propose a plug-and-play Semantic Probe framework consisting of an EAR module and a dynamic BAS loss, which revives in-domain fine-tuning methods and achieves SOTA performance on four CDFSL benchmarks.
^† indicates equal contribution (co-first authors).

CVPR 2026 Workshop

Jiacong Liu, Shu Luo, Yikai Qin, Yaze Zhao, Yongwei Jiang, Yixiong Zou

We propose GiPL, an efficient two-branch training framework for CDFSOD. In the first branch, we design an iterative pseudo-label self-training paradigm, which performs zero-shot inference on the support set to generate reliable pseudo-annotations, fuses them with ground-truth labels, and iteratively optimizes the model. In the second branch, we introduce generative data augmentation pipeline using large vision-language models, which synthesizes domain-aligned, multi-object annotated images.

CVPR 2026 Workshop

CDiscover team from HUST

As part of NTIRE 2026, we hosted the second CD-FSOD Challenge to systematically evaluate and promote progress in detecting objects in unseen target domains under limited annotation conditions. The challenge received strong community interest, with 128 registered participants and a total of 696 submissions. Participants explored a wide range of strategies, introducing innovative methods that push the performance frontier under both open-source and closed-source tracks. This report presents a detailed overview of the NTIRE 2026 CD-FSOD Challenge.

CVPR 2026

Yaze Zhao, Yixiong Zou, Yuhua Li, Ruixuan Li

This work first identifies and addresses the local feature misalignment problem in CDFSL scenarios based on vision-language models (CLIP). To tackle this issue, we propose the CC-CDFSL method, which constructs self-supervised signals via bidirectional feature transformation and introduces the Semantic Anchor mechanism to mitigate noise interference in the visual modality.

2026.03 CVPR Workshop NTIRE-CDFSOD Challenge (Cross‑Domain Few‑Shot Object Detection) – Global Second Place
2025.05 Chinese Collegiate Computing Competition (4C), Central South Regional Third Prize
2024.08 RAICOM Robot Developer Competition, National First Prize
2024.06 Blue Bridge Cup National Software and Information Technology Talent Competition, National Second Prize
2024.06 China College Students’ Service Outsourcing Innovation and Entrepreneurship Competition, National Second Prize
2024.05 Mathematical Contest in Modeling (MCM), Honorable Mention (H Award)
2024.04 Huawei ICT Competition (Ascend AI Track), National Third Prize

2025.10 First‑Class Master’s Academic Scholarship, HUST
2025.06 HUST Outstanding Graduate
2025.05 HUST Outstanding Communist Youth League Member
2024.12 HUST Merit Student
2024.03 Sangfor Scholarship
2023.12 Academic Excellence Scholarship, HUST
2023.07 Outstanding Individual in the “Three Rural” Summer Social Practice, HUST
2022.03 Social Welfare Scholarship, HUST

2021.06, Lorem ipsum dolor sit amet, consectetur adipiscing elit. Vivamus ornare aliquet ipsum, ac tempus justo dapibus sit amet.
2021.03, Lorem ipsum dolor sit amet, consectetur adipiscing elit. Vivamus ornare aliquet ipsum, ac tempus justo dapibus sit amet. | [video]