2025秋季学期
2026.01.08
- 卢昕怡 Gains: Fine-grained Federated Domain Adaptation in Open Set [paper][slides]
2025.12.24
- 蒋明忠
COIDO Efficient Data Selection for Visual Instruction Tuning via Coupled Importance-Diversity Optimization[paper]
SegEarth-OV: Towards Training-Free Open-Vocabulary Segmentation for Remote Sensing Images[paper]
SegEarth-OV3 Exploring SAM 3 for Open-Vocabulary Semantic Segmentation in Remote Sensing Images[paper]
all[slides]
2025.12.17
- 郑腾鑫陵 PRISM: PROGRESSIVE ROBUST LEARNING FOR OPEN-WORLD CONTINUAL CATEGORY DISCOVERY[paper1]HuangmeiSinger A Dataset and A Branchformer-Diffusion Model for Huangmei Opera Synthesis[paper2][slides]
paper1 连续类别发现（CCD）旨在利用在已知类别上训练的模型，从连续到达的未标记数据流中自动发现新的类别概念，同时保留识别先前已知类别的能力。尽管最近取得了进展，但现有的方法通常假设所有阶段的数据都来自一个单一的平稳分布——在开放世界场景中很少满足这一条件。在本文中，通过引入开放世界连续类别发现（OW-CCD）设置来挑战这种平稳分布假设。
paper2 歌声合成在元宇宙、音乐创作和娱乐、文化保护和传承中得到了广泛的应用。然而，由于缺乏专业注释的高质量数据集和适当的深度学习模型，黄梅戏等传统戏曲的合成受到了限制。作者开发了一个演唱声音数据集，并提出了一个为黄梅戏独特演唱风格量身定制的声学模型。
2025.12.10
- 张文良 Self-Expansion of Pre-trained Models with Mixture of Adaptersfor Continual Learning [paper] [slides]
- 孙佳家 on-the-importance-of-language-driven-representation-learning-for-heterogeneous-federated-learning-Paper-Conference [paper] [slides]
2025.12.03
- 郑金鹏 Test-time Adaptation on Graphs via Adaptive Subgraph-based Selection and Regularized Prototypes [paper][slides]
2025.11.26
- 卢昕怡 CLIP-driven Coarse-to-fine Semantic Guidance for Fine- grained Open-set Semi-supervised Learning & OSLOPROMPT: Bridging Low-Supervision Challenges and Open-Set Domain Generalization in CLIP [paper1] [paper2] [slides]
- 郑腾鑫陵 Maintaining Consistent Inter-Class Topology in Continual Test-Time Adaptation[paper][slides]
作者通过构建类一致性拓扑图，从类间关系、类内关系、批不平衡拓扑加权三个维度解决持续TTA的重点挑战：错误累积。
2025.11.19
- 王远见 microCLIP: Unsupervised CLIP Adaptation via Coarse-Fine Token Fusion for Fine-Grained Image Classification[paper][slides]
2025.11.12
- 蒋明忠Active Learning for Open-set Annotation[paper][slides]
- 王子晨 Language-Guided Transformer for Federated Multi-Label Classification[paper][slides]
2025.11.05
- 孙佳家 Mixture of Experts Made Personalized：Federated Prompt Learning for Vision-Language Models[paper][slides]
2025.10.22
- 张文良 Mind the Gap Preserving and Compensating for the Modality Gap in CLIP-Based Continual Learning[paper][slides]
2025.10.15
- 卢昕怡 Adaptive Test-Time Personalization for Federated Learning & FedCTTA: A Collaborative Approach to Continual Test-Time Adaptation in Federated Learning [paper1] [paper2] [slides]
第一篇文章提出测试时个性化联邦学习场景和自适应测试时个性化方法，通过从源客户端，之间的分布偏移中，自适应学习模型每个模块的适配率，为应对多样分布偏移提供灵活性。第二篇文章是持续TTA的场景，通过基于随机生成噪声样本的模型输出分布计算相似性感知聚合，在保障数据隐私的同时实现自适应知识共享。
- 郑腾鑫陵 Mixture-of-Experts for Open Set Domain Adaptation A Dual-Space Detection Approach[paper][slides]
2025.9.24
- 王远见 MLC-NC: Long-Tailed Multi-Label Image Classification Through the Lens of Neural Collapse[paper][slides]
2025.9.17
- 蒋明忠 DiffCLIP: Few-shot Language-driven Multimodal Classifier[paper][slides]
- 王子晨 Focal-SAM: Focal Sharpness-Aware Minimization for Long-Tailed Classification[paper][slides]
2025.9.10
- 孙佳家 Federated Learning with Domain Shift Eraser [paper][slides]
- 张文良 Rethinking Query-based Transformer for Continual Image Segmentation[paper][slides]