The paper introduces MuPHI, a diagnostic benchmark that isolates multimodal harm arising solely from image‑text composition, providing benign counterfactuals and annotated harm rationales. Evaluation with MuPHI shows that vision‑language models can ground individual modalities but struggle to bind evidence into a coherent harm mechanism, and that existing harm detection methods transfer poorly to new harm distributions. To address this, the authors propose Compositional Harm Interpretation Learning (CHIL), which rewards evidence identification, cross‑modal binding, and verdict consistency, leading to significant improvements in out‑of‑distribution transfer and counterfactual robustness.
By Anisha Saha, Varsha Suresh, Teodora Kamova, Sophia Wiedmann, Timothy Hospedales, Vera Demberg
Large vision-language models (LVLMs) have recently shown immense potential in automated content moderation, sparking growing interest in developing harmful-video benchmarks. However, we identify two primary limitations in existing works: 1) The multi-layered characteristics of harmful videos are overlooked.
The paper introduces MPS-Bench, a benchmark of 5,181 scenarios from 584 real-world images across 12 high-risk domains, each paired with a hidden user profile, to evaluate personalized safety in vision‑language models (VLMs). Eight leading VLMs were tested and found to almost always respond directly (86‑99%) without seeking missing context, scoring no higher than 2.6/5 on personalized safety. The authors identify a phenomenon called visual dominance, where visual information enters text representations early and suppresses textual risk signals, and propose PRISM, a lightweight input monitor that predicts when a query should be deferred, achieving 0.978 AUC and outperforming all tested models on the safety‑utility Pareto frontier.
By Edward Sun, Yuchen Wu, Zixian Ma, Eric Hanchen Jiang, Yijia Xiao, Xiaoyuan Yi, Ranjay Krishna, Wei Wang, Jindong Wang, Aylin Caliskan
arXiv:2609.06991v1 Announce Type: cross
Abstract: Recent text-to-audio-video (T2AV) models jointly generate video, speech, sound effects, and ambience from a single text prompt. This capability poses...
By Suah Choi, Tae-Young Lee, Gyeong-Moon Park
Multi2AV‑Safety is a new benchmark for evaluating safety in multimodal-to-audio‑video generation. It covers all 11 non‑singleton conditioning configurations (text, image, audio, video) and contains 11,024 attack instances. The benchmark reveals that safety guards often fail when harmful semantics arise from combinations of benign inputs or when explicit harmful cues are masked by benign multimodal context, highlighting a gap in compositional risk perception.
By Kaichao Jiang, Changtao Miao, Baiqi Wu, Zhiyuan Lu, Kang Yang, Peiwei Zhao, Junchi Chen, Yunfeng Diao, He Liu, Qi Chu, Tao Gong, Nenghai Yu
The paper presents a new taxonomy for online harm on video platforms, expanding existing frameworks to multimodal contexts and defining six content-based categories: Information, Hate and harassment, Addictive, Clickbait, Sexual, and Physical harms. It also evaluates multimodal large language models (MLLMs) as alternative annotators, comparing GPT‑4‑Turbo, non‑trained crowdworkers, and trained human annotators on a dataset of 19,422 potentially harmful YouTube videos. The study finds that MLLMs can perform comparably to human annotators in both binary and multi‑label classification tasks, suggesting potential applications in content moderation and social science research.
By Claire Wonjeong Jo, Miki Weso{\l}owska, Magdalena Wojcieszak
arXiv:2606. 02111v1 Announce Type: cross Abstract: As multimodal large language models (MLLMs) have advanced to process video inputs, concerns have emerged about their potential for malicious misuse.
By Choongwon Kang, Seungjong Sun, Hyunmin Jun, Jang Hyun Kim
arXiv:2606. 02443v1 Announce Type: cross Abstract: Between the first visible sign of danger and the moment an accident occurs, there is often a window where intervention remains possible.
By Yusong Zhao, Yuejin Xie, Youliang Yuan, Junjie Hu, Jitian Guo, Yujiu Yang, Pinjia He
arXiv:2609.20850v1 Announce Type: new
Abstract: While Multimodal Large Language Models (MLLMs) show remarkable advancements, their cross-modal capabilities introduce complex vulnerabilities that easi...
By Yueming Lyu, Yilian Shi, Haoxiang Tan, Linzhuang Zou, Qihao Wang, Guihua Yu, Jie Qin, Xin Gao, Chenyang Si, Jing Dong, Caifeng Shan
arXiv:2606. 05177v1 Announce Type: cross Abstract: Existing multimodal safety benchmarks focus solely on visual inputs and cannot assess Omni Large Language Models (LLMs) that process vision, audio, and text.
By Manh Luong, Tamas Abraham, Junae Kim, Amar Kaur, Rollin Omari, Gholamreza Haffari, Trang Vu, Lizhen Qu, Dinh Phung
arXiv:2608. 00076v2 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) increasingly support high-stakes decision making by combining complementary information from images and text.
By Vahidin Hasic, Chao Wang, Luis C. Garcia-Peraza-Herrera, David Watson, Senka Krivic
arXiv:2608. 06865v1 Announce Type: cross Abstract: The malicious use of generative artificial intelligence to create highly realistic deepfake videos raises serious ethical concerns and poses substantial challenges to AI safety.
By Xuechao Zou, Shun Zhang, Kai Li, Yi Zhou, Xinyu Sun, Yuhui Chen, Zhe Wu, Congyan Lang, Junliang Xing