Two researchers, one shared question. 两位研究者，同一个问题。

Tianjian Zhou is a Ph.D. researcher in the continuous Bachelor–Master–Ph.D. program at the National University of Defense Technology, in the College of Systems Engineering and the Laboratory for Big Data and Decision. His research spans self-supervised representation learning, efficient adaptation of frozen foundation models and 360° perception; the common thread is how much a visual model can learn from fewer labels, fewer trained parameters and less complete views.

He is first author of EquiSSL (ACM TOG '26 / SIGGRAPH Asia 2026 Journal Track), DAGA (NeurIPS '26), CORE (IEEE TNNLS '26) and SDI (Pattern Recognition '26), and works on geometric deep learning, self-supervised representation learning and diffusion models. He maintains Visual SSL Digest, a daily reading list of self-supervised and geometric vision papers.

周天健是国防科技大学系统工程学院、大数据与决策实验室的博士研究生，就读于本硕博连读项目。他的研究涵盖自监督表征学习、冻结基础模型的高效适配与 360° 感知；共同的问题是：视觉模型能否用更少的标签、更少的训练参数和更不完整的视角，学到有用的结构。

他是 EquiSSL（ACM TOG '26 / SIGGRAPH Asia 2026 期刊轨）、DAGA（NeurIPS '26）、CORE（IEEE TNNLS '26）与 SDI（Pattern Recognition '26）的第一作者，研究几何深度学习、自监督表征学习与扩散模型。他维护 Visual SSL Digest，每日整理自监督学习与几何视觉方向的论文。

Yishan Li is a Ph.D. researcher and collaborator at NUDT. Her work sits where representation learning meets multi-modal perception: infrared–visible fusion, contrastive manifold autoencoders and equivariant 360° vision. The three settings share a question — what should a representation preserve when the input changes — and she is most interested in answers that survive a change of sensor, spectrum or camera pose.

She is a co-author of EquiSSL (ACM TOG '26) and DAGA (NeurIPS '26), co-first author of CMAE (IEEE ICME 2026, Oral), and a co-author of CORE, SDI, HTNet and TriCM-Net. Alongside these projects she works on cross-modal retrieval and on fusion architectures that do not assume the modalities are symmetric.

李易珊是国防科技大学博士研究生、团队合作研究者。她的研究位于表征学习与多模态感知的交汇处：红外—可见光融合、对比流形自编码与等变 360° 视觉。这三类场景共享同一个问题——当输入发生变化时，表征应当保留什么——她尤其关注那些在传感器、成像波段或相机位姿改变后依然成立的答案。

她是 EquiSSL（ACM TOG '26）与 DAGA（NeurIPS '26）的合作者、CMAE（IEEE ICME 2026，Oral）的共同第一作者，并参与 CORE、SDI、HTNet 与 TriCM-Net 的研究。此外，她也研究跨模态检索，以及不假定模态对称的融合架构。