MLSP-P28: Robustness and Security in Large Language Models II
Poster
Wed, 6 May, 14:00 - 16:00
Location: Poster Area 6
Session Type: Poster
Session Chair: Xiangui Kang, Sun Yat-Sen University
Track: Machine Learning for Signal Processing [ML]
Click the to view the manuscript on IEEE Xplore Open Preview
 

MLSP-P28.1: JAILFUZZ: A ZERO-KNOWLEDGE AND STATE FEEDBACK BASED GRAY-BOX FUZZING FRAMEWORK FOR LARGE LANGUAGE MODELS
Huan Liu, University of Science and Technology of China, China; Xiaoshan Sun, Yingjun Zhang, Yang Zhang, Institute of Software Chinese Academy of Sciences, China
 

MLSP-P28.2: Dual-Criterion Sample Selection for Noisy Labels: Integrating Neighborhood Prediction Divergence and Loss Values
Qian Rong, Lu Zhang, Ling Yuan, Guohui Li, China, China
 

MLSP-P28.3: HIBAR: A HIDDEN BACKDOOR ATTACK ON LLM RECOMMENDATION SERVICES VIA MULTI-TURN DIALOGUE MANIPULATION
Tianbo Zhang, Tongwei Liu, Luoyu Mei, Tao Guo, Aibo Song, Southeast University, China
 

MLSP-P28.4: SAFEGRAD: GRADIENT SURGERY FOR SAFE LLM FINE-TUNING
Biao Yi, Nankai University, China; Jiahao Li, Xi'an Jiaotong University, China; Baolei Zhang, Lihai Nie, Tong Li, Nankai University, China; Tiansheng Huang, Georgia Institute of Technology, United States of America; Zheli Liu, Nankai University, China
 

MLSP-P28.5: DINN: KEY-ACTIVATED MODEL STEGANOGRAPHY WITH DYNAMIC SPARSE INVERTIBLE NEURAL NETWORKS
Shimian Fang, Sun Yat-sen University, China; Jibin Zheng, Foshan University, China; Xiangui Kang, Sun Yat-sen University, China; Sheng Li, Fudan University, China; Jane Z Wang, University of British Columbia, Canada; Yan Li, Sun Yat-sen University Cancer Center, China
 

MLSP-P28.6: DHEval: A Dynamic Hallucination Evaluation Protocol Robust to Data Contamination
Jinhua Yin, Tsinghua University, China; Du Li, University of Electronic Science and Technology of China, China; Wendan Wang, Beijing University of Posts and Telecommunications, China; Peiru Yang, Huili Wang, Yunlong Ma, Yongfeng Huang, Tsinghua University, China; Tao Qi, Beijing University of Posts and Telecommunications, China
 

MLSP-P28.7: FUZZY MEMBERSHIP-ENHANCED UNCERTAINTY-AWARE FUSION FOR MULTI-VIEW CLASSIFICATION
Zijian Tang, Yuqing Hou, Xuelei He, Xiaowei He, Fengjun Zhao, NorthWest University, China
 

MLSP-P28.8: ORSC: OBJECT-AWARE REINFORCEMENT WITH SEMANTIC CONSISTENCY FOR HALLUCINATION MITIGATION IN MLLMS
Jinyang He, Xiaoyu Shi, University of Chinese Academy of Sciences, China; Hong Xie, University of Science and Technology of China, China; Yan Zhang, Mingsheng Shang, University of Chinese Academy of Sciences, China
 

MLSP-P28.9: STRUCTURE-AWARE ADVERSARIAL PURIFICATION: DYNAMIC MASKING AND ATTRIBUTION REFINEMENT IN DIFFUSION MODELS
Chaochao Fan, Wei Lu, Cyberspace Institute of Advanced Technology, Guangzhou University, China; Denghui Zhang, Cyberspace Institute of Advanced Technology, Guangzhou University; China and Key Laboratory of Computing Power Network and Information Security, Ministry of Education; Qilu University of Technology (Shandong Academy of Sciences), China; Dawei Zhao, China and Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Qilu University of Technology (Shandong Academy of Sciences), China; Zhaoquan Gu, School of Computer Science and Technology, Harbin Institute of Technology (Shenzhen), China
 

MLSP-P28.10: PAST AS PRIOR: REWEIGHTED PROXY GUIDANCE FOR STABLE ADVERSARIAL TRAINING
Yaohua Liu, The University of Hong Kong, China; Jiaxin Gao, The Hong Kong Polytechnic University, China