苏修，男，中南大学特聘教授，博士生导师。先后入选 国家级高层次青年人才，第十六批湖南省“百人计划”人才。博士毕业于悉尼大学，师从Chang Xu教授。在国际知名CCF/CAAI A类会议和期刊 TPAMI、CVPR、NeurIPS、ICML、ICCV、AAAI、KDD、ACMMM、ECCV、ICLR等发表论文40余篇，长期担任相关顶级会议和期刊的审稿人和程序委员会委员。个人研究内容主页：https://xiusu.github.io/。

【研究方向】具身智能、世界模型、多模态大模型（医学）、计算机视觉、自动化机器学习等。实验室长期专注于人工智能模型架构设计、多模态学习方法、具身机器人控制等领域，有深厚的人工智能算法和软硬件积累。欢迎感兴趣的同学和合作者联系。

【招生信息】博士、硕士、RA（过渡出国、读博）、本科生。课题组氛围融洽，与澳大利亚和香港众多顶尖名校深度合作，且国外顶尖名校长期有名额，每年多个入学季，欢迎有兴趣的同学加入(xiusu1994@csu.edu.cn)。2026年 目前组内剩余入学名额： 2 港科博士，1 中南博士，0 中南硕士。

【组内优势】 1. 硬件资源丰富。有充足的显卡等计算资源以及机器人（机械臂、灵巧手）资源，目前组内有 宇树、松灵、UR、因时 的 人形机器人、轮式机器人，包含 机械臂和机器手。

                     2. 尊重学生的个人选择。无论是学术界或企业界发展，包括：出国留学、大学任教或企业工作，都会提供充足的指导和资源支持。

                     3. 组内有众多 海外/国内 名校学生。学术氛围浓厚，能及时跟踪相关领域前沿技术和发展趋势。

                     4. 与众多海外名校和国内知名AI企业保持长期深度科研合作关系。定期输送优秀人才去海外QS 前50学校读博/研 和 国内知名AI企业。 

                     5. 课题组经费充足，能保证科研所需的实验条件，并提供细致的科研指导和充足的助研津贴。

【代表性论文】（详见 https://xiusu.github.io/）

---

[1] Large Language Models Driven Neural Architecture Search for Universal and Lightweight Disease Diagnosis on Histopathology Slide Images (Digital Medicine, Nature Partner Journals)

[2] BCNetV2: Searching for Network Width With Bilaterally Coupled Network. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI, CCF A)

[3] Injection without Distortion: Geometrically Constrained Knowledge Enhancement for Vision-Language Models (AAAI, CCF A, Oral), 2025

[4] Multi-Modal Style Transfer-based Prompt Tuning for Efficient Federated Domain Generalization (AAAI, CCF A, Oral), 2025

[5] ROVER: Robust Generative Continual Identity Unlearning against Relearning Attacks (AAAI, CCF A), 2025

[6] UtilGen: Utility-Centric Generative Data Augmentation with Dual-Level Task Adaptation (NeurIPS, CCF A), 2025

[7] L-MTP: Leap Multi-Token Prediction Beyond Adjacent Context for Large Language Models (NeurIPS, CCF A), 2025

[8] On the Stability and Generalization of Meta-Learning: the Impact of Inner-Levels (NeurIPS, CCF A), 2025

[9] Modeling Inter-Gaussian Mutual Information for Dynamic Novel View Synthesis (ACMMM, CCF A), 2025

[10] Graph Unlearning Meets Influence-aware Negative Preference Optimization (ACMMM, CCF A), 2025

[11] Addressing Granularity-induced Semantic Drift in OvOD via Graph-guided semantically consistent representation (ACMMM, CCF A), 2025

[12] DualFPT: Handling Data Heterogeneity in Federated Prompt Tuning from both Generalized and Personalized Perspective (ACMMM, CCF A), 2025

[13] Identify, Isolate, and Purge: Mitigating Hallucinations in LVLMs via Self-Evolving Distillation (ACMMM, CCF A), 2025

[14] CounterPC: Counterfactual Feature Realignment for Unsupervised Domain Adaptation on Point Clouds (ICCV, CCF A), 2025

[15] Stable Fair Graph Representation Learning with Lipschitz Constraint. International Conference on Machine Learning (ICML, CCF A), 2025

[16] TinyMIG: Transferring Generalization from Vision Foundation Models to Single-Domain Medical Imaging. International Conference on Machine Learning (ICML, CCF A), 2025

[17] VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR, CCF A, Oral), 2025

[18] Harmonizing for defect visibility with Fine-Grained Hierarchical Interaction Learning. International Conference on Acoustics, Speech and Signal Processing (ICASSP, CCF B), 2025

[19] HieClip: Hierarchical CLIP with Explicit Alignment for Zero-Shot Anomaly Detection. International Conference on Acoustics, Speech and Signal Processing (ICASSP, CCF B), 2025

[20] Perturbating, Tuning, and Collaborating: Harnessing Vision Foundation Models for Single Domain Generalization on Medical Imaging. AAAI Conference on Artificial Intelligence (AAAI, CCF A), 2025

[21] Seeing Beyond Noise: Joint Graph Structure Evaluation and Denoising for Multimodal Recommendation. AAAI Conference on Artificial Intelligence (AAAI, CCF A), 2025

[22] Universal Frequency Domain Perturbation for Single-Source Domain Generalization. ACM Multimedia (ACM MM, CCF A), 2024

[23] Detecting Any Human-Object Interaction Relationship: Universal HOI Detector with Spatial Prompt Learning on Foundation Models. Advances in Neural Information Processing Systems (NeurIPS, CCF A), 2024

[24] Detecting Any Instruction-to-Answer Interaction Relationship:Universal Instruction-to-Answer Navigator for Med-VQA. International Conference on Machine Learning (ICML, CCF A), 2024

[25] BEYOND THE LIMIT OF WEIGHT-SHARING: PIONEERING SPACE-EVOLVING NAS WITH LARGE LANGUAGE MODELS. International Conference on Acoustics, Speech, and Signal Processing (ICASSP, CCF B), 2024

[26] TCNAS: TRANSFORMER ARCHITECTURE EVOLVING IN CODE CLONE DETECTION. International Conference on Acoustics, Speech, and Signal Processing (ICASSP, CCF B), 2024

[27] PROMPTING LABEL EFFICIENCY IN FEDERATED GRAPH LEARNING VIA PERSONALIZED SEMI-SUPERVISION. International Conference on Acoustics, Speech, and Signal Processing (ICASSP, CCF B), 2024

[28] DiffNAS: Bootstrapping Diffusion Models by Prompting for Better Architectures. IEEE International Conference on Data Mining (ICDM, CCF B), 2023

[29] Re-mine, Learn and Reason: Exploring the Cross-modal Semantic Correlations for Language-guided HOI detection. International Conference on Computer Vision (ICCV, CCF A), 2023

[30] Neural Architecture Search for Wide Spectrum Adversarial Robustness. AAAI Conference on Artificial Intelligence (AAAI, CCF A), 2023

[31] Searching for Better Spatio-temporal Alignment in Few-Shot Action Recognition. Conference and Workshop on Neural Information Processing Systems (NeurIPS, CCF A), 2022

[32] Sufficient Vision Transformer. ACM SIGKDD Conference on Knowledge Discovery and Data Mining (SIGKDD, CCF A), 2022

[33] Vision Transformer Architecture Search. European Conference on Computer Vision (ECCV, CAAI A), 2022

[34] ScaleNet: Searching for the Model to Scale. European Conference on Computer Vision (ECCV, CAAI A), 2022

[35] K-shot NAS: LearnableWeight-Sharing for NAS with K-shot Supernets. International Conference on Machine Learning (ICML, CCF A), 2021

[36] BCNet: Searching for Network Width with Bilaterally Coupled Network. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR, CCF A), 2021

[37] Prioritized Architecture Sampling with Monto-Carlo Tree Search. The IEEE Conference on Computer Vision and Pattern Recognition (CVPR, CCF A), 2021

[38] Locally Free Weight Sharing for Network Width Search. International Conference on Learning Representations (ICLR, CCF A, Spotlight), 2021

[39] Data Agnostic Filter Gating for Efficient Deep Networks. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP, CCF B), 2022

[40] Automatic bridge crack detection using a convolutional neural network. Applied Sciences, 2019

[41] An efficient hole-filling method based on depth map in 3D view generation. 2017 International Conference on Optical Instruments and Technology: Optoelectronic Imaging/Spectroscopy and Signal Processing Technology

[42] An improved three-dimension reconstruction method based on guided filter and Delaunay. 2017 International Conference on Optical Instruments and Technology: Optoelectronic Imaging/Spectroscopy and Signal Processing Technology