Shenghua Liu

Professor, AI for Trustworthy Foundation Models and Agents

sliu.png

No.6 Kexueyuan South Road, Haidian District

Beijing, China 100190

[short bio] [group page]

email: liushenghua at ict.ac.cn

I am a Professor at Institute of Computing Technology, Chinese Academy of Sciences. My research focuses on two closely related directions:

  • AI for trustworthy foundation models and agents in the era of RSI, with an emphasis on building reliable, faithful, and capable foundation models and agents for scientific research.
  • Graph intelligence and network mining, including agent reasoning over complex relational structures, and graph-structured organization of agent memory, as well as anomaly detection in real-world networks.

The featured works are published on IEEE TKDE, ACM TKDD, and proceedings of top-tier conferences such as AAAI, ICLR, ACL, CIKM, WSDM, ECML-PKDD, etc. Some of the publications are recognized as ASP-DAC 2010 best paper candidate, ECML-PKDD 2020 best student DM paper award.

My educational and visiting experience:

news

Apr 16, 2026 Four of our works are accepted by ACL 2026: two in the main conference and two in findings. Two papers are accepted by ICML 2026 (one Spotlight). Two papers are accepted by ECCV 2026.
May 16, 2025 Two of our works are accepted by ACL main 2025.

selected publications

  1. Prism-Δ: Differential Subspace Steering for Prompt Highlighting in Large Language Models
    Yuyao Ge, Shenghua Liu, Yiwei Wang, and 6 more authors
    In Conference on Language Modeling, COLM, 2026
    PDF Code Website
  2. Gated Differentiable Working Memory for Long-Context Language Modeling
    Lingrui Mei, Shenghua Liu, Yiwei Wang, and 7 more authors
    In Proc. of the Association for Computational Linguistics, ACL Main, Jul 2026
    DOI PDF
  3. HiddenGuard: Fine-Grained Safe Generation with Specialized Representation Router
    Lingrui Mei, Shenghua Liu, Yiwei Wang, and 3 more authors
    In Proc. of the Association for Computational Linguistics, ACL Main, Jul 2026
    DOI PDF
  4. a1: Steep Test-time Scaling Law via Environment Augmented Generation
    Lingrui Mei, Shenghua Liu, Yiwei Wang, and 5 more authors
    In Proc. of the Association for Computational Linguistics, ACL Findings, Jul 2026
    DOI PDF
  5. Reward and Guidance through Rubrics: Promoting Exploration to Improve Multi-Domain Reasoning
    Baolong Bi, Shenghua Liu, Yiwei Wang, and 6 more authors
    In Proc. of the International Conference on Machine Learning, ICML, Jul 2026
    Spotlight
    arXiv PDF
  6. Who is in the Spotlight: The Hidden Bias Undermining Multimodal Retrieval-Augmented Generation
    Jiayu Yao, Shenghua Liu, Yiwei Wang, and 5 more authors
    In Proc. of the Empirical Methods in Natural Language Processing, EMNLP Main, Nov 2025
    PDF
  7. Context-DPO: Aligning Language Models for Context-Faithfulness
    Baolong Bi, Shaohan Huang, Yiwei Wang, and 11 more authors
    In Proc. of the Association for Computational Linguistics, ACL Findings, Jul 2025
    DOI PDF
  8. Decoding by Contrasting Knowledge: Enhancing Large Language Model Confidence on Edited Facts
    Baolong Bi, Shenghua Liu, Lingrui Mei, and 4 more authors
    In Proc. of the Association for Computational Linguistics, ACL Main, Jul 2025
    DOI PDF
  9. Can Graph Descriptive Order Affect Solving Graph Problems with LLMs?
    Yuyao Ge, Shenghua Liu, Baolong Bi, and 5 more authors
    In Proc. of the Association for Computational Linguistics, ACL Main, Jul 2025
    DOI PDF
  10. Is Factuality Enhancement a Free Lunch For LLMs? Better Factuality Can Lead to Worse Context-Faithfulness
    Baolong Bi, Shenghua Liu, Yiwei Wang, and 5 more authors
    In International Conference on Learning Representations, ICLR, Jul 2025
    PDF
  11. "Not Aligned" is Not "Malicious": Being Careful about Hallucinations of Large Language Models’ Jailbreak
    Lingrui Mei, Shenghua Liu, Yiwei Wang, and 3 more authors
    In Proc. of the International Conference on Computational Linguistics, Coling, Jul 2025
    PDF
  12. SLANG: New Concept Comprehension of Large Language Models
    Lingrui Mei, Shenghua Liu, Yiwei Wang, and 2 more authors
    In Proc. of the Empirical Methods in Natural Language Processing, EMNLP Main, Jul 2024
    PDF
  13. Node Embedding Preserving Graph Summarization
    Houquan Zhou, Shenghua Liu, Huawei Shen, and 1 more author
    ACM Transactions on Knowledge Discovery from Data, TKDD, Jul 2024
    PDF
  14. Graph Summarization for Preserving Spectral Characteristics
    Houquan Zhou, Shenghua Liu, Huawei Shen, and 1 more author
    In Proc. of the SIAM International Conference on Data Mining, SDM, Jul 2024
    PDF
  15. Unified Dense Subgraph Detection: Fast Spectral Theory based Algorithms
    Wenjie Feng, Shenghua Liu, Danai Koutra, and 1 more author
    IEEE Transactions on Knowledge and Data Engineering, TKDE, Jul 2024
    Published March 2024 (pub date: 17 July 2023)
    PDF
  16. Time Series Anomaly Detection with Adversarial Reconstruction Networks
    Shenghua Liu, Bin Zhou, Quan Ding, and 4 more authors
    IEEE Transactions on Knowledge and Data Engineering, TKDE, Jul 2023
    PDF Code
  17. SpecGreedy: Unified Dense Subgraph Detection
    Wenjie Feng, Shenghua Liu, Danai Koutra, and 2 more authors
    In Proc. of the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML-PKDD, Jul 2020
    Best student DM paper award. Acceptance rate: 19%. Verified on 40 real-world networks, and a 1.47-billion-edge graph
    PDF Code
  18. A Contrast Metric for Fraud Detection in Rich Graphs
    Shenghua Liu, Bryan Hooi, and Christos Faloutsos
    IEEE Transactions on Knowledge and Data Engineering, TKDE, Jul 2019
    PDF
  19. HoloScope: Topology-and-Spike Aware Fraud Detection
    Shenghua Liu, Bryan Hooi, and Christos Faloutsos
    In Proc. of the ACM International Conference on Information and Knowledge Management, CIKM, Jul 2017
    PDF Code Slides