写在前面： 在人工智能新时代，我们致力于培养具备批判性思维和创新能力的学生，帮助他们解决已知与未知的问题。我们鼓励学生保持对物理世界与数字世界的持续好奇心，勇于提出新问题，培养他们成为大数据与人工智能领域的未来领军人才。这将帮助学生更好地适应数字化、智能化的新时代，推动产业智能化转型与创新变革。

当前，我们专注于通用人工智能（AGI）的探索，致力于让机器智能具备与人类相似的思考、规划、推理与行动能力，使人工智能无处不在，深刻融入我们的日常生活与工作，推动社会的全面智能化。

研究方向：聚焦数据科学与人工智能前沿技术

· AI智能体方向：聚焦AI智能体的自主决策、持续学习与自我进化，尤其在复杂任务拆解方面，致力于实现从“被动执行”到“主动规划”的转变。深入研究语言与环境交互、多模态融合等核心技术，为AGI的实现提供算法支持，推动智能体在多样化复杂场景中的自主适应与高效执行。场景聚焦：AI for Science（科学领域）-自主AI科学家、生命科学发现、以及金融领域

· 大语言模型方向：研究大模型的高效训练、轻量化部署和场景适配，解决“大模型泛化能力不足、落地成本高、场景适配性弱”等问题。推动大模型在各行业垂直领域的定制化落地，构建高性能AI基础设施，助力AGI的迭代升级与产业数字化赋能。

· 大模型数据方向：聚焦大模型训练与优化所需的核心数据支持，重点研究数据采集、清洗、高质量筛选、数据标注标准化、隐私保护以及增量数据更新等技术，解决数据质量不均、隐私风险高、模型与数据适配性不足的问题，建立高质量、安全可复用的数据资源体系，为大模型高效训练与精准优化提供数据保障。

· 智算方向：构建符合AGI需求的高效节能智算体系，重点研究智算架构创新、算力调度优化与能效提升等关键技术，解决算力瓶颈与数据处理低效问题。推动智算与AI技术的深度融合，为语言智能体、大模型及数据处理提供稳定、高效、低成本的算力支持，助力AGI低成本、规模化发展与产业落地。

Google Scholar：https://scholar.google.com/citations?user=N4rQYDAAAAAJ&hl=en

欢迎具有相关专业知识背景的优秀学生报考，Email是直接联系到我最有效的方式。

博士，毕业于日本东京大学，美国加州大学伯克利分校联合培养。长期专注于通用人工智能（AGI）及AI基础设施的前沿研究，旨在赋予机器类人的思考、规划与自主决策能力，以技术创新推动产业的智能化升级。深耕学术界与工业界多年，曾任职于中国科学院软件研究所、奇虎360搜索事业部、阿里巴巴达摩院及阿里云智能集团，具备带领大规模产品与研发团队的丰富技术创新与产业落地经验。在学术领域，已在KDD、NeurIPS、SIGIR、WWW、CVPR、ICCV、ECCV、AAAI、ICDE、ACL、TPAMI、TOIS、TKDE等数据科学与人工智能领域的顶级会议和期刊上发表论文百余篇。曾多次主持国家自然科学基金（面上/青年）、教育部留学回国人员科研启动基金及北京市自然科学基金等项目，推动前沿科研与产业应用的深度融合。

部分科研成果：

• Evo-Retriever: LLM-Guided Curriculum Evolution with Viewpoint-Pathway Collaboration for Multimodal Document Retrieval. CVPR 2026
• Co-EPG: A Framework for Co-Evolution of Planning and Grounding in Autonomous GUI Agents. AAAI 2026
• Importance-aware data selection for efficient llm instruction tuning. AAAI 2026
• RASD: Retrieval -Augmented Speculative Decoding, ACL 2025
• AirCache:  Activating Inter-modal Relevancy KV Cache Compression for Efficient Vision-Language Model Inference, ICCV 2025
• DBCopilot: Natural Language Querying over Massive Databases via Schema Routing, EBDT 2025 (Best Paper Runner-Up)
• Match, Compare, or  Select? An Investigation of Large Language Models for Entity Matching,  COLING 2025
• ChartM3: A Multi-Stage Code-Driven Pipeline for Constructing Multi-Dimensional and Multi-Step Visual Reasoning Data in Chart Comprehension. EMNLP 2025
• Self-Paced  Unified Representation Learning for Hierarchical Multi-Label Classification. AAAI 2024
• Adaptive quantization error reconstruction for llms with mixed precision, COLM 2024
• Mixture-of-LoRAs:  An  Efficient Multitask Tuning for Large Language Models, COLING 2024
• Federated unlearning for on-device recommendation, ACM WSDM 2023
• UUKG: Unified  Urban  Knowledge  Graph Dataset  for  Urban Spatiotemporal Prediction. NeurIPS 2023
• A Good  Student is Cooperative and Reliable: CNN- Transformer Collaborative Learning for Semantic Segmentation. ICCV 2023
• A Knowledge-Enhanced Framework for Imitative Transportation Trajectory Generation. ICDM 2022
• Federated learning for personalized humor recognition, ACM Transactions on Intelligent Systems and Technology. ACM TIST, 2022
• Keyword-Guided Neural  Conversational Model,  The  Thirty- Fifth AAAI Conference on Artificial Intelligence. AAAI 2021
• CARE:  Commonsense-Aware Emotional Response Generation with Latent Concepts. AAAI 2021
• Adapting to Context-Aware Knowledge in Natural Conversation for Multi-Turn Response Selection. WWW 2021
• Initialization Matters: Regularizing Manifold-informed Initialization for Neural Recommendation Systems. The 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining. KDD 2021
• Learning to Ask Appropriate Questions in Conversational Recommendation. The 44th International ACM SIGIR Conference on Research and    Development in Information Retrieval. SIGIR 2021
• Relational  Learning with Gated and Attentive Neighbor Aggregator for Few-Shot Knowledge Graph Completion. The 44th International ACM SIGIR Conference on Research and Development in Information Retrieval. SIGIR 2021
• Uniting heterogeneity, inductiveness, and efficiency for graph representation learning. IEEE Transactions on Knowledge and Data Engineering. IEEE TKDE, 2021
• MusicBERT: A self-supervised learning of music representation. The 29th ACM International Conference on Multimedia (ACM MM 2021)
• Crsal: Conversational recommender systems with adversarial learning. ACM Transactions on Information Systems (ACM TOIS), 2020
• Towards Persona-Based Empathetic Conversational Models. EMNLP 2020
• Next point-of-interest recommendation on resource-constrained mobile devices. WWW 2020
• Online sales prediction via trend alignment-based multitask recurrent neural networks. Knowledge and Information Systems, 2020
• Machine Comprehension-Incorporated Relevance Matching. ICDM 2019
• Semi-supervised domain adaptation via Fredholm integral based kernel methods. Pattern Recognition, 2019
• Enhancing collaborative filtering with generative Augmentation.   The 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD 2019)
• Resumevis: A visual analytics system to discover semantic information in semi- structured resume data. ACM Transactions on Intelligent Systems and Technology (ACM TIST), 2018
• Multi-view Adversarially Learned  Inference for Cross-domain Joint Distribution Matching. KDD 2018
• Redundancy-resistant Generative Hashing for Image Retrieval. IJCAI 2018
• Tada: trend alignment with dual- attention multi-task recurrent neural networks for sales prediction. ICDM 2018
• Emotion analysis for personality inference from EEG signals. IEEE transactions on affective computing (IEEE TAC), 2018
• PME: projected metric embedding on heterogeneous networks for link prediction. KDD 2018
• Stockassistant: a stock ai assistant for reliability modeling of stock comments. KDD 2018
• Semi-supervised deep generative modelling of incomplete   multi-modality emotional data. Proceedings of the 26th ACM international conference on Multimedia (ACM MM 2018)
• Neural memory streaming recommender networks with adversarial training. Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (KDD 2018)
• A machine reading comprehension-based approach for featured snippet extraction. ICDM 2018
• Information-theoretic domain adaptation under severe noise conditions. AAAI 2018
• Discriminative transformation for multi-dimensional temporal sequences. IEEE Transactions on Image Processing (TIP), 2017
• A Location- Sentiment-Aware Recommender System for Both Home-Town and Out-of-Town Users. KDD2017
• Spatial-aware hierarchical collaborative deep learning for POI recommendation. IEEE Transactions on Knowledge and Data Engineering (TKDE), 2017
• Fredholm multiple kernel learning for semi-supervised domain adaptation. AAAI 2017
• An automatic approach for transit advertising in public transportation systems. ICDM 2017
• Discriminative dimensionality reduction for multi- dimensional sequences. IEEE transactions on pattern analysis and machine intelligence (TPAMI), 2017
• A hybrid term–term relations analysis approach for topic detection. Knowledge-Based Systems, 2016
• Discovering Interpretable Geo-Social Communities for User Behavior Prediction. The 32nd IEEE International Conference on Data Engineering, ICDE 16
• Hierarchical dynamic parsing and encoding for action recognition. ECCV 2016
• Learning graph-based poi embedding for location-based recommendation. Proceedings of the 25th ACM international on conference on information and knowledge management (CIKM 2016)
• Transfer Feature Representation via Multiple Kernel Learning. AAAI 2015
• Cross-Domain Metric Learning Based on Information Theory. AAAI 2014