Award Papers
We are proud to announce the following papers that have been recognized with awards at the AI-Safety Workshop IJCAI 2026. Congratulations to all the award-winning authors!
-
Best Paper Award
From Echo to Nexus: How Attacker Personality Dictates Success in Jailbreaking LLMs
Yuhang Wang, Jian Zhao, Rubin He, Huilin Zhou, Tianle Zhang, Rui Feng and Xuelong Li -
Best Student Paper Award
LatentGuard: Safeguarding Autonomous Driving VLMs via Latent World Logic Repair
Tianyuan Zhang, Maoran Ye, Yang Qu, Jiangfan Liu, Zonglei Jing, Taichuan Li, Aishan Liu, Jiakai Wang, Yuqing Ma and Xianglong Liu -
Best Paper Nomination Award
SREM: Semantic-space Reconstruction Evidence Mining for Generalizable AI-Generated Image Detection
Anwei Luo, Jinsong Hu, Yi Tu, Chenqi Kong and Dan Ma -
Best Paper Nomination Award
Hydra: Towards Massively Multi-Target Backdoor Attacks against VLMs
Siyuan Liang, Zhantao Yang, Yiming Li, Tong Zhang, Shangwen Zhu, Xiujin Liu and Dacheng Tao -
Best Paper Nomination Award
Forensic-Aware Continual Adaptation with Fisher-Guided LoRA Surgery for Image Forgery Localization
Chenqi Kong, Song Xia, Anwei Luo and Xuanang Cheng
Accepted Papers
Below are the accepted papers for the AI-Safety Workshop IJCAI 2026. Click each link to view or download the full PDF.
-
Metadata-Borne Malicious Instructions: Hiding Agent Tool-Abuse Payloads in Image Metadata for LLM Agent Security Evaluation
Chen Ma, Shuaidong Li, Junhao Zheng and Jing Li -
Cross-Modal Decoupled Attacks on Autonomous LLM Agents
Jun Zhang, Yuming Liu, Huilin Zhou, Jian Zhao, Tianle Zhang, Wenqi Ren and Xuelong Li -
Contextual Permission Boundary Testing in the OpenClaw Security Attack and Defense Challenge: Task-Only Red-Teaming via Execution Traces
Zeren Luo, Jingyi Zheng, Yule Liu, Zhen Sun and Xinlei He -
Task–Skill Consistency, Skill-Carried Authorization Context, and Trace-Guided Repair in the OpenClaw Security Challenge
Zhongjian Wang -
The Instruction Gate: Work-Order Camouflage and the Bottleneck in Layered LLM Agent Defenses
Hang Ding, Qi Liu, Aoxu Ji, Ruihao Xu and Mao Luo -
Multi-Channel Workflow Framing in Tool-Using Agent Security: A Case Study from OpenClaw
Yinuo Xu -
Scope-Gap Injection: Exploiting Evaluator Channel Blind Spots in the OpenClaw Agent Security Competition
Weihang Niu, Mengxing Ren and Che Qu -
When Approval Becomes Execution: Authorization-Grounded Workflow Attacks on Tool-Using Agents
Chenggong Shi and Hongmei Mou -
An Empirical Case Study of Prompt Engineering Strategies for Bypassing Multi-Layer Agent Security Defenses
Chengwei Chu -
AutoAuthScan: Detecting Self-Service Authorization Risks in Tool-Using Agents
Yixiang Zhang, Zhipeng Li and Haojie Yuan -
Less is More: Modality-Decoupling for General AIGC Audio-Video Detection
Jielun Peng, Yaqi Li, Yabin Wang, Jincheng Liu, Xiaopeng Hong and Athanasios V. Vasilakos -
Modality-Specific Latent Realness Modeling for General AI-Generated Audio-Video Detection
Haoran Chen, Daiqi Gao, Tianzheng Wang, Mingming Deng, Yirong Sun, Jianfeng Dong and Xun Wang -
GEF-VA: Gated Evidence Fusion for General AI-Generated Video-Audio Detection
Shuaibo Li, Laixin Zhang, Weilin Ruan, Hongqiu Wang, Jialu Li, Lei Zhu and Wei Ma -
Sparse Visual Expert Activation and Structured Fusion for AI-Generated Audio-Video Detection
Haoyu Wang, Haohao Li, Hongshuo Jin, Junhao Wang, Qing Wen, Kui Ren, Zhongjie Ba, Peng Cheng, Li Lu and Zhan Qin -
MVADetector: Modality-Specific Manifold Contrastive Learning for General AIGC Audio-Video Detection
Yuning Zhang, Tingyu Liu, Mingyu Liao and Xinghao Wang -
PATE-Forensics: Perception-as-Tool for Explainable Deepfake Forensics with General-Purpose MLLMs
Yaqi Li, Jielun Peng, Yabin Wang, Jincheng Liu and Xiaopeng Hong -
A Reproducible Detection, Localization, and Explanation System for DDL-X Track 3
Yihang Lu, Tianshuo Zhang, Siran Peng, Haoyuan Zhang and Zhen Lei -
Deepfake Detection, Localization, and Trace Explanation with Multi-Task Visual Modeling
Xing Fan and Jiayue Yu -
TRACE: Token Reference Attention for Auditable Deepfake Region Localization
Yi-Fang Wang, You-Cheng Zhao, Chih-Yu Jian, Chia-Ming Lee and Chih-Chung Hsu -
OpenSDI-based Deepfake Detection and Localization for the IJCAI 2026 DDL-X Challenge Track 3
Zhekai Wang -
Anchored-prompt Zero-Training Framework for Deepfake Detection, Localization and Explanation
Xibo Fan, Junbo Gao, Wenfei Xu and Ye Zhu -
TRACE: Trajectory-Guided Attack Material Synthesis for Tool-Using Agents
Yingkun Huang and Xiaoru Zhuang -
TRACE: An Evidence-Grounded Benchmark for Safety Evaluation of Large Reasoning Models
Zhenyu Wu, Siyuan Chen, Changchun Yang, Jiaqi Dong, Min Zhou, Ali Almadan, Talal Hammad, Faisal Wahbo, Aminullah Tora, Mona Alshahrani and Xin Gao -
SOP-Inject: Cross-Resource Procedure Disguise and Authorization Misbinding in Tool-Using Agents
Ruoyu Chen, Yang Hong, Zihan Zhao and Wenbo Hu