个人简介
石海龙,博士,中国科学院微电子研究所青年研究员,AI中心具身智能组负责人,博士生导师,中国科学院高层次人才引进,博士毕业于中国科学院计算技术研究所,前阿里巴巴集团智能零售算法团队负责人、京东集团IoT事业部算法团队负责人。研究方向包括计算机视觉、多模态大模型、具身智能、群体智能等,主持或作为技术负责人承担过多项中国科学院先导专项、国家科技重大专项、973、国家自然科学基金等项目,在IEEE/ACM Transactions和CCF-A等期刊会议上发表论文40余篇,申请发明专利20余项。
研究方向
- 具身智能:研究VLA、VLN、世界模型等具身智能关键技术,核心工作是把“从真实场景到仿真训练再到真机部署”的全链路自动化,已完成端到端Pipeline,并在园区安防巡检等场景进行应用验证,相关成果荣获阿里巴巴最具商业价值奖。
- 多模态大模型:跨模态数据融合、语义理解与知识表示,基于Agent的多模态内容理解与生成,领域大模型微调与后训练等,发表TCSVT/TIST/MM等顶会&顶刊多篇。
- 大模型推理加速:研究多模态大模型推理加速技术,包括模型架构设计、模型压缩剪枝量化、软硬协同优化等,发表CVPR/TIOT等多篇论文。
- 计算机视觉: 复杂环境(低光照、雨雪、遮挡)下的图像识别、检测、跟踪与增强,发表ECCV/AAAI/IJCAI/TIFS 等顶会&顶刊多篇。
- 数据挖掘与检索: 大规模信息检索、推荐、广告系统,CTR/CVR 预估,工业级亿级用户系统。
工作经历
2023.07—至今 中国科学院微电子研究所,人工智能芯片与系统研发中心,博士生导师,中国科学院高层次人才计划,具身智能课题组负责人
- 围绕计算机视觉、具身智能、多模态大模型、大模型推理加速等方向深入研究:
- 发表20余篇顶会&顶刊论文,申请发明专利7项。
2019.11—2023.06,阿里巴巴集团,智能零售算法团队负责人
- 负责阿里智能零售算法研发与团队管理,主导零售推荐、营销增长、智能定价、流量调控等近10项大型项目的研发;
- 支持数亿用户的实时商品与商家推荐,以及数十万商家、数亿商品的智能化运营。
2018.03—2019.11,京东集团,IoT事业部算法团队负责人
- 负责IoT事业部算法研发与团队管理,主导智慧门店与智能对话等系统研发;
- 支撑京东数百家直营门店、数万家便利店的智能运营,数千万智能设备的交互控制;
- 多次受邀在Datafun、InfoQ等行业峰会上做主题报告。
2015.01—2018.02,中国科学院计算技术研究所,助理研究员
- 在国家科技专项、院先导等支持下,围绕物联网基础软硬件、协议与系统开展研究;
- 完成2款物联网智能芯片、10余款物联网节点、近10个物联网应用系统研发;
- 累计发表论文18篇,申请专利7项,软著2项。
进三年代表性成果
会议论文
- [KDD 2026] Bocheng Pan, Hailong Shi*, Xingyu Gao. HiST: Hierarchical Semantic Tree Augmentation for Generative Recommendation, The 32nd SIGKDD Conference on Knowledge Discovery and Data Mining, Jeju, South Korea, August 9-13, 2026. CCF-A
- [CVPR 2026] Tong Xu, Hailong Shi*, Xingyu Gao. SCoRe: Salience-Coverage Reduction for Vision Token Pruning in Vision-Language Models, The IEEE/CVF Conference on Computer Vision and Pattern Recognition, Denver, USA, June 3-7, 2026. CCF-A
- [ACL 2026] Zhuoning Zhu, Xingyu Gao, Hailong Shi*. MENTOR: Mitigating Identity Drift in Dynamic Role-Playing via Dual-Chain Structured Memory, The 64th Annual Meeting of the Association for Computational Linguistics, San Diego, California, July 2-7, 2026.
- [ACM MM 2025] Bocheng Pan, Hailong Shi*, Xingyu Gao. DR-VQA: Decompose-then-Reconstruct for Visual Question Answering in BLV Assistance, ACM International Conference on Multimedia, Dublin, Ireland, 27th-31st October, 2025. CCF-A
- [ICML 2025] Wanjin Feng, Xingyu Gao, Wenqian Du, Hailong Shi, Peilin Zhao, Pengcheng Wu, Chunyan Miao. Efficient Parallel Training Methods for Spiking Neural Networks with Constant Time Complexity, Forty-second International Conference on Machine Learning, Vancouver, Canada, 13th-19th July, 2025. CCF-A
- [IJCAI 2024] Huan Li, Hailong Shi*, Xingyu Gao. A Coarse-To-Fine Fusion Network for Event-Based Image Deblurring, The 33rd International Joint Conference on Artificial Intelligence, Jeju, South Korea, 3th-9th August, 2024. CCF-A
- [ECCV 2024] Qi Guo, Hailong Shi*, Huan Li, Jinsheng Xiao, Xingyu Gao. REDIR: Refocus-free Event-based De-occlusion Image Reconstruction, The 18th European Conference on Computer Vision, MiCo Milano, Italy, Sep 29th-Oct 4th, 2024. 计算机视觉三大顶会之一
- [ECIR 2024] Luo Ji#, Jiayu Mao#, Hailong Shi*, Qian Li, Hongxia Yang. An Adaptive Framework of Geographical Group-Specific Network on O2O Recommendation, The 46th European Conference on Information Retrieval, Glasgow, Scotland, 24th-28th March, 2024.
- [ICDM 2023] Wanjin Feng, Hailong Shi, Peilin Zhao, Xingyu Gao. Mixtron: Bandit Online Multiclass Prediction with Implicit Feedback, International Conference on Data Mining, Shanghai, China, 1th-4th December, 2023.
期刊论文
- [ACM TIST] Qi Guo, Xingyu Gao, Hailong Shi*. Spatiotemporal-Guided Training-Free Urban Partial Vision Restoration, ACM Transactions on Intelligent Systems and Technology, Accepted. 影响因子6.6
- [ACM TIST] Xingyu Gao, Bocheng Pan, Zhenyu Chen, Baobin Zhang, Fei Sun, Hailong Shi*. Multi-Modal Collaboration Evaluation of Large Language Models via Image Captioning, ACM Transactions on Intelligent Systems and Technology, Accepted. 影响因子6.6
- [ACM TOMM] Jiefan Qiu, Dongfu Zhu, Xingyu Gao, Mengqi Jiang, Jiahan Song, Hailong Shi*. mmWave Radar-based Personalized Multi-object Vital Signs Monitoring, ACM Transactions on Multimedia Computing, Communications, and Applications, 2026. 影响因子6.0
- [IEEE TIFS] Jinsheng Xiao, Hao Ma, Ruidi Chen, Xingyu Gao, Hailong Shi*, Zhongyuan Wang. STKPS-Net:Spatio-Temporal Key Patch Selection Network for Anomalous Action Recognition, IEEE Transactions on Information Forensics and Security, 2026. 中科院一区TOP,影响因子8.0,CCF-A
- [ACM TIOT] Bocheng Pan, Hailong Shi*, Xingyu Gao. MPVCD: Multi-Perspective Visual Contrastive Decoding for Reliable Assistance, ACM Transactions on Internet of Things, 2026. 影响因子3.7
- [IEEE TCSVT] Xingyu Gao, Zuolei Li, Hailong Shi*, Zhenyu Chen, Peilin Zhao. Scribble-Supervised Video Object Segmentation via Scribble Enhancement, IEEE Transactions on Circuits and Systems for Video Technology, 2025. SCI一区,影响因子8.3
- [ACM TOMM] Shaojun Zhu, Bincheng Zhu, Kaikai Chi, Jiefan Qiu, Hailong Shi, and Xingyu Gao. Maximizing Long-Term Task Completion Ratio of UAV-Enabled Wirelessly-Powered MEC Systems, ACM Transactions on Multimedia Computing, Communications, and Applications, 2025. SCI一区,影响因子5.2
- [IEEE TNNLS] Hang Ran, Xingyu Gao, Lusi Li, Weijun Li, Songsong Tian, Gang Wang, Hailong Shi, Xin Ning. Brain-Inspired Fast- and Slow-Update Prompt Tuning for Few-Shot Class-Incremental Learning, IEEE Transactions on Neural Networks and Learning Systems, 2024. SCI一区,影响因子10.2
教育经历
- 2010.09—2015.01 中国科学院计算技术研究所,博士
- 2008.09—2010.06 武汉大学,电子信息学院,硕士
- 2004.09—2008.06 武汉大学,电子信息学院,本科
