Gongshen Liu
Tenured Professor
Works on AI security, large language models and their safety, and natural language processing.
Gongshen Liu's lab at SJTU works on large language models and interpretability, evaluation & safety of LMs. 52 papers since 2023, 18 at top venues, mostly Findings of ACL and CVPR. Applications close .
At a glance
- papers, 2023–2026
- 14/ year
- at T1 venues
- 35%
- PhD students
- —
- MS students
- —
- alumni listed
- —
Papers by yearfrom DBLP, 2023–2026
- 12 papers2026 · T1 6 · collecting
- 18 papers2025 · T1 5
- 13 papers2024 · T1 5
- 9 papers2023 · T1 2
T1 venuesOtherVenues: Findings of ACL, CVPR, ICASSP, ACL, IJCNN
Research focuspaper titles that use each topic's words
- Large language models15 papers
- language11 papers
- security3 papers
- Interpretability, evaluation & safety of LMs2 papers
- safety1 paper
Paper list
From DBLP.
202612 papers
- Int. J. Approx. Reason.A robust multi-source transfer classification method based on belief functions for cross-domain pattern recognition (opens in a new tab)
- CVPR (T1)Consensus Entropy: Harnessing Multi-VLM Agreement for Self-Verifying and Self-Improving OCR (opens in a new tab)
- KSEMDomain Adaptation of MLLM-Based GUI Agents in Documentation-Rich Environments with Standard Operating Procedures (opens in a new tab)
- Findings of ACLEVA: Evolving Semantic Adversaries for Red-Teaming GUI Agents Against Environmental Injection Attacks (opens in a new tab)
- ICLR (T1)GhostEI-Bench: Do Mobile Agent Resilience to Environmental Injection in Dynamic On-Device Environments? (opens in a new tab)
- Mach. Intell. Res.Incomplete Data Classification via Distribution Alignment With Evidence Combination (opens in a new tab)
- CVPR (T1)LaSM: Layer-wise Scaling Mechanism for Defending Pop-up Attack on GUI Agents (opens in a new tab)
- ICMROpenImplicit: Benchmarking Implicit Reasoning in MLLMs via Open-Ended Evaluation (opens in a new tab)
- ACL (T1)SEARL: Joint Optimization of Policy and Tool Graph Memory for Self-Evolving Agents (opens in a new tab)
- CVPR (T1)See, Think, Act: Teaching Multimodal Agents to Effectively Interact with GUI by Identifying Toggles (opens in a new tab)
- Findings of ACLThe Confidence Paradox: Unveiling the Latent Discriminative Power of Diffusion Large Language Models in Mathematical Reasoning (opens in a new tab)
- CVPR (T1)Training High-Level Schedulers with Execution-Feedback Reinforcement Learning for Long-Horizon GUI Automation (opens in a new tab)
202518 papers
- COLING (T1)ALIS: Aligned LLM Instruction Security Strategy for Unsafe Input Prompt (opens in a new tab)
- IEEE Trans. Neural Networks Learn. Syst.Backdoor Attacks and Countermeasures in Natural Language Processing Models: A Comprehensive Security Review (opens in a new tab)
- HP3CBi-mark: A copyright marking framework for AI-generative-text (opens in a new tab)
- ICCV (T1)Can Knowledge be Transferred from Unimodal to Multimodal? Investigating the Transitivity of Multimodal Knowledge Editing (opens in a new tab)
- IEEE Trans. Neural Networks Learn. Syst.FedSTS: A Stratified Client Selection Framework for Consistently Fast Federated Learning (opens in a new tab)
- COLING (T1)Gracefully Filtering Backdoor Samples for Generative Large Language Models without Retraining (opens in a new tab)
- SMCGTRNet: Graph Topology-aware Refinement Network for User Role Classification in Social Networks (opens in a new tab)
- Findings of EMNLPHidden Ghost Hand: Unveiling Backdoor Vulnerabilities in MLLM-Powered Mobile GUI Agents (opens in a new tab)
- ACM Comput. Surv.Igniting Language Intelligence: The Hitchhiker's Guide from Chain-of-Thought Reasoning to Language Agents (opens in a new tab)
- NeurocomputingImproving semi-autoregressive machine translation with the guidance of syntactic dependency parsing structure (opens in a new tab)
- ICMEKMoP: Knowledge-injected Mixture-of-Prefix for Joint Multimodal Aspect-Based Sentiment Analysis (opens in a new tab)
- IEEE Trans. Veh. Technol.MKF-ADS: Multi-Knowledge Fusion Based Anomaly Detection System in Vehicular Control Area Networks (opens in a new tab)
202413 papers
- ACL (T1)Acquiring Clean Language Models from Backdoor Poisoned Datasets by Downscaling Frequency Space (opens in a new tab)
- LREC/COLINGBackdoor NLP Models via AI-Generated Text (opens in a new tab)
- IEEE Trans. Instrum. Meas.Convolutional Feature Aggregation Network With Self-Supervised Learning and Decision Fusion for SAR Target Recognition (opens in a new tab)
- ICANNFashionGPT: A Large Vision-Language Model for Enhancing Fashion Understanding (opens in a new tab)
- LREC/COLINGHow Large Language Models Encode Context Knowledge? A Layer-Wise Probing Study (opens in a new tab)
- NLPCCImproving Non-autoregressive Machine Translation with Error Exposure and Consistency Regularization (opens in a new tab)
- ACL (T1)Investigating Multi-Hop Factual Shortcuts in Knowledge Editing of Large Language Models (opens in a new tab)
- USENIX Security (T1)Less Defined Knowledge and More True Alarms: Reference-based Phishing Detection without a Pre-defined Reference List (opens in a new tab)
- Peer Peer Netw. Appl.LSF-IDM: Deep learning-based lightweight semantic fusion intrusion detection model for automotive (opens in a new tab)
- ICASSP (T1)Multi-Grained Multimodal Interaction Network for Sentiment Analysis (opens in a new tab)
- ICASSP (T1)NWS: Natural Textual Backdoor Attacks Via Word Substitution (opens in a new tab)
- Findings of EMNLPR-Judge: Benchmarking Safety Risk Awareness for LLM Agents (opens in a new tab)
20239 papers
- Future Gener. Comput. Syst.DESC-IDS: Towards an efficient real-time automotive intrusion detection system based on deep evolving stream clustering (opens in a new tab)
- ICASSP (T1)FedPrompt: Communication-Efficient and Privacy-Preserving Prompt Tuning in Federated Learning (opens in a new tab)
- Findings of ACLIs Continuous Prompt a Combination of Discrete Prompts? Towards a Novel View for Interpreting Continuous Prompts (opens in a new tab)
- IJCNNNeural Linguistic Steganography with Controllable Security (opens in a new tab)
- AAAI (T1)PLMmark: A Secure and Robust Black-Box Watermarking Framework for Pre-trained Language Models (opens in a new tab)
- IJCNNRobust Secret Data Hiding for Transformer-based Neural Machine Translation (opens in a new tab)
- IJCNNSCS-VAE: Generate Style Headlines Via Novel Disentanglement (opens in a new tab)
- ICONIPSDPSAT: Syntactic Dependency Parsing Structure-Guided Semi-Autoregressive Machine Translation (opens in a new tab)
- ICANNSemantic Information Mining and Fusion Method for Bot Detection (opens in a new tab)
Where alumni went
Alumni are not listed on the lab homepage.
Apply to SJTUFall 2027
Computer Science PhD (international applicants)
- Round
- Self-funded
- Opens
- —
- GRE
- See program page
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- DETUM
Matthias Grabmair
10 T1 - GBEdinburgh
Helen Hastie
- Reasoning & agents
- CNSJTU
Pan Hu
- Large language models
- USStanford
Daniel E. Ho
- Large language models
- USUCSD
Ndapa Nakashole
- Knowledge, QA & retrieval
- USUMD
Mohit Iyyer
- Large language models
Your notes
Save this lab to keep notes.
Data
Some numbers could not be verified:
- The member list could not be found on the lab site.
- The alumni list could not be found on the lab site.
- Collected
- Verified
- Papers
- DBLP
- Members
- Not collected
- Alumni
- Not collected
Tiers are venue classifications, not lab rankings.
SJTUcloses