Researcher — paper.dinq.me

2026 顶会论文

作者位次 #2/5ICML 2026Poster· Social Aspects / Robustness

Contrastive Spectral Rectification: Test-Time Defense towards Zero-shot Adversarial Robustness of CLIP

Sen Nie · Jie Zhang · Zhuo Wang · Shiguang Shan · Xilin Chen

安全社会与人本OpenReview

作者位次 #5/6ICML 2026Poster· Deep Learning / Robustness

You Don't Protect if You Don't Expect: Breaking the Key Assumption behind CLIP's Test-Time Defenses

Ruize Zhang · Yu Li · Zhang Wan · Juan Cao · Jie Zhang · Sheng Tang

OpenReview

作者位次 #2/5CVPR 2026Poster· Computer Vision & Pattern Recognition

V-Attack: Targeting Disentangled Value Features for Controllable Adversarial Attacks on LVLMs

Sen Nie · Jie Zhang · Jianxin Yan · Shiguang Shan · Xilin CHEN

安全多模态OpenReview 项目页

作者位次 #3/5ACL 2026Poster· Natural Language Processing

INFACT : A Diagnostic Benchmark for Induced Faithfulness and Factuality Hallucinations in Video- LLM s

Yangjunqi · Yuecong Min · Jie Zhang · Shiguang Shan · Xilin CHEN

摘要

Despite rapid progress, Video Large Language Models (Video-LLMs) remain unreliable due to hallucinations, which are outputs that contradict either video evidence (faithfulness) or verifiable world knowledge (factuality).Existing benchmarks provide limited coverage of factuality hallucinations and...

多模态评测与基准检索与知识OpenReview 项目页

Jie Zhang

获取完整研究档案

联系方式

2026 顶会论文

Contrastive Spectral Rectification: Test-Time Defense towards Zero-shot Adversarial Robustness of CLIP

You Don't Protect if You Don't Expect: Breaking the Key Assumption behind CLIP's Test-Time Defenses

V-Attack: Targeting Disentangled Value Features for Controllable Adversarial Attacks on LVLMs

INFACT : A Diagnostic Benchmark for Induced Faithfulness and Factuality Hallucinations in Video- LLM s

常见合作者