Cover Image

When Safe Unimodal Inputs Collide: Optimizing Reasoning Chains for Cross-Modal Safety in Multimodal Large Language Models

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Cai, Wei, Liu, Shujuan, Zhao, Jian, Shi, Ziyan, Zhao, Yusheng, Yuan, Yuchen, Zhang, Tianle, Zhang, Chi, Li, Xuelong
Format:	Preprint
Published:	2025
Subjects:	Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Safe Semantics, Unsafe Interpretations: Tackling Implicit Reasoning Safety in Large Vision-Language Models
by: Cai, Wei, et al.
Published: (2025)

A Parameter-Efficient Mixture-of-Experts Framework for Cross-Modal Geo-Localization
by: Li, LinFeng, et al.
Published: (2025)

Visual Attention Reasoning via Hierarchical Search and Self-Verification
by: Cai, Wei, et al.
Published: (2025)

TeleAI-Safety: A comprehensive LLM jailbreaking benchmark towards attacks, defenses, and evaluations
by: Chen, Xiuyuan, et al.
Published: (2025)

Aetheria: A multimodal interpretable content safety framework based on multi-agent debate and collaboration
by: He, Yuxiang, et al.
Published: (2025)

When Modalities Conflict: How Unimodal Reasoning Uncertainty Governs Preference Dynamics in MLLMs
by: Zhang, Zhuoran, et al.
Published: (2025)

Metis: Learning to Jailbreak LLMs via Self-Evolving Metacognitive Policy Optimization
by: Zhou, Huilin, et al.
Published: (2026)

RADAR: A Risk-Aware Dynamic Multi-Agent Framework for LLM Safety Evaluation via Role-Specialized Collaboration
by: Chen, Xiuyuan, et al.
Published: (2025)

Beyond Unimodal Shortcuts: MLLMs as Cross-Modal Reasoners for Grounded Named Entity Recognition
by: Ma, Jinlong, et al.
Published: (2026)

Multimodality Helps Unimodality: Cross-Modal Few-Shot Learning with Multimodal Models
by: Lin, Zhiqiu, et al.
Published: (2023)

SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilities
by: Jiang, Fengqing, et al.
Published: (2025)

A Systematic Study of Cross-Modal Typographic Attacks on Audio-Visual Reasoning
by: Chen, Tianle, et al.
Published: (2026)

StitchFusion: Weaving Any Visual Modalities to Enhance Multimodal Semantic Segmentation
by: Li, Bingyu, et al.
Published: (2024)

U3M: Unbiased Multiscale Modal Fusion Model for Multimodal Semantic Segmentation
by: Li, Bingyu, et al.
Published: (2024)

Segment-Aligned Policy Optimization for Multi-Modal Reasoning
by: Gao, Lei, et al.
Published: (2026)

Seeing is Fixing: Cross-Modal Reasoning with Multimodal LLMs for Visual Software Issue Fixing
by: Huang, Kai, et al.
Published: (2025)

Multimodal Chain-of-Thought Reasoning in Language Models
by: Zhang, Zhuosheng, et al.
Published: (2023)

Dark Side of Modalities: Reinforced Multimodal Distillation for Multimodal Knowledge Graph Reasoning
by: Zhao, Yu, et al.
Published: (2025)

Beyond Unimodal Boundaries: Generative Recommendation with Multimodal Semantics
by: Zhu, Jing, et al.
Published: (2025)

Rethink the Role of Deep Learning towards Large-scale Quantum Systems
by: Zhao, Yusheng, et al.
Published: (2025)

Rebalanced Multimodal Learning with Data-aware Unimodal Sampling
by: Jiang, Qingyuan, et al.
Published: (2025)

Multimodal Representation Learning by Alternating Unimodal Adaptation
by: Zhang, Xiaohui, et al.
Published: (2023)

From Perception to Reasoning: Deep Thinking Empowers Multimodal Large Language Models
by: Zhu, Wenxin, et al.
Published: (2025)

Boosting Multimodal Federated Learning via Chained Modality Optimization
by: Zhang, Zixin, et al.
Published: (2026)

ERF-BA-TFD+: A Multimodal Model for Audio-Visual Deepfake Detection
by: Zhang, Xin, et al.
Published: (2025)

MARCO: Meta-Reflection with Cross-Referencing for Code Reasoning
by: Zhao, Yusheng, et al.
Published: (2025)

UMCL: Unimodal-generated Multimodal Contrastive Learning for Cross-compression-rate Deepfake Detection
by: Lai, Ching-Yi, et al.
Published: (2025)

When Safety Fails Before the Answer: Benchmarking Harmful Behavior Detection in Reasoning Chains
by: Kakkar, Ishita, et al.
Published: (2026)

Understanding Unimodal Bias in Multimodal Deep Linear Networks
by: Zhang, Yedi, et al.
Published: (2023)

On the Value of Cross-Modal Misalignment in Multimodal Representation Learning
by: Cai, Yichao, et al.
Published: (2025)

MAGE: Multimodal Alignment and Generation Enhancement via Bridging Visual and Semantic Spaces
by: E, Shaojun, et al.
Published: (2025)

Reading, Not Thinking: Understanding and Bridging the Modality Gap When Text Becomes Pixels in Multimodal LLMs
by: Sun, Kaiser, et al.
Published: (2026)

Decoupling the Image Perception and Multimodal Reasoning for Reasoning Segmentation with Digital Twin Representations
by: Li, Yizhen, et al.
Published: (2025)

When Safety Collides: Resolving Multi-Category Harmful Conflicts in Text-to-Image Diffusion via Adaptive Safety Guidance
by: Xiang, Yongli, et al.
Published: (2026)

Meta-Learn Unimodal Signals with Weak Supervision for Multimodal Sentiment Analysis
by: Mai, Sijie, et al.
Published: (2024)

Missing Modality Prediction for Unpaired Multimodal Learning via Joint Embedding of Unimodal Models
by: Kim, Donggeun, et al.
Published: (2024)

SafeMT: Multi-turn Safety for Multimodal Language Models
by: Zhu, Han, et al.
Published: (2025)

When One Modality Sabotages the Others: A Diagnostic Lens on Multimodal Reasoning
by: Zhang, Chenyu, et al.
Published: (2025)

D-CAT: Decoupled Cross-Attention Transfer between Sensor Modalities for Unimodal Inference
by: Daher, Leen, et al.
Published: (2025)

Membership Inference for Contrastive Pre-training Models with Text-only PII Queries
by: Cheng, Ruoxi, et al.
Published: (2026)