mlsanityData to 14 Oct 2026

Cross-Modal Safety Mechanism Transfer in Large Vision-Language Models

arXiv 2410.12662 ↗First posted 16 Oct 2024cs.CV cs.AI cs.CLcited by 19

On the map: Multimodal LLMs / vision-language models → MLLM jailbreaks and adversarial robustness

Abstract (arXiv)

Vision-language alignment in Large Vision-Language Models (LVLMs) successfully enables LLMs to understand visual input. However, we find that existing vision-language alignment methods fail to transfer the existing safety mechanism for text in LLMs to vision, which leads to vulnerabilities in toxic image. To explore the cause of this problem, we give the insightful explanation of where and how the safety mechanism of LVLMs operates and conduct comparative analysis between text and vision. We find that the hidden states at the specific transformer layers play a crucial role in the successful activation of safety mechanism, while the vision-language alignment at hidden states level in current methods is insufficient. This results in a semantic shift for input images compared to text in hidden states, therefore misleads the safety mechanism. To address this, we propose a novel Text-Guided vision-language Alignment method (TGA) for LVLMs. TGA retrieves the texts related to input vision and uses them to guide the projection of vision into the hidden states space in LLMs. Experiments show that TGA not only successfully transfers the safety mechanism for text in basic LLMs to vision in vision-language alignment for LVLMs without any safety fine-tuning on the visual modality but also maintains the general performance on various vision tasks (Safe and Good).

Key facts. Cited by 19 papers in the map (data through 14 Oct 2026). Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 30 of its 37 references that are in the map. First posted on arXiv 16 Oct 2024.

Builds on

30 of 37 references.

Closest ancestors · 4

Same direction, newest first

  1. 2024A Survey of Attacks on Large Vision-Language Models: Resources, Advances, and Future Trends10 Jul
  2. 2024Cross-Modal Safety Alignment: Is textual unlearning all you need?27 May
  3. 2024Eyes Closed, Safety On: Protecting Multimodal LLMs via Image-to-Text Transformation14 Mar
  4. 2023FigStep: Jailbreaking Large Vision-Language Models via Typographic Visual Prompts9 Nov

Other references · 26

  1. 2017Attention Is All You Needcited by 60,361
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  3. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  4. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  5. 2023Visual Instruction Tuningcited by 7,869
  6. 2023Qwen Technical Reportcited by 3,344
  7. 2023Mistral 7Bcited by 3,164
  8. 2022Locating and Editing Factual Associations in GPTcited by 2,691
  9. 2023Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyondcited by 2,224
  10. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answeringcited by 1,919
  11. 2023Evaluating Object Hallucination in Large Vision-Language Modelscited by 1,729
  12. 2021LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairscited by 1,555
14 more
  1. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilitiescited by 1,239
  2. 2021Conceptual 12M: Pushing Web-Scale Image-Text Pre-Training To Recognize Long-Tail Visual Conceptscited by 1,216
  3. 2019BERT Rediscovers the Classical NLP Pipelinecited by 1,059
  4. 2023SEED-Bench: Benchmarking Multimodal LLMs with Generative Comprehensioncited by 940
  5. 2021Knowledge Neurons in Pretrained Transformerscited by 601
  6. 2022Image as a Foreign Language: BEiT Pretraining for All Vision and Vision-Language Taskscited by 511
  7. 2023Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructionscited by 387
  8. 2023DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Modelscited by 376
  9. 2024Safety Fine-Tuning at (Almost) No Cost: A Baseline for Vision Large Language Modelscited by 167
  10. 2023Search-in-the-Chain: Interactively Enhancing Large Language Models with Search for Knowledge-intensive Taskscited by 73
  11. 2023ToViLaG: Your Visual-Language Generative Model is Also An Evildoercited by 27
  12. 2024Unsupervised Information Refinement Training of Large Language Models for Retrieval-Augmented Generationcited by 19
  13. 2023HOD: A Benchmark Dataset for Harmful Object Detectioncited by 10
  14. 2024A Theory for Token-Level Harmonization in Retrieval-Augmented Generationcited by 7

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 1; 2025: 11; 2026: 7.

19 papers.

Sort:

By area:

  1. 2025A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deploymentcited by 116
  2. 2024Cross-modal Information Flow in Multimodal Large Language Modelscited by 67
  3. 2025A Survey of Safety on Large Vision-Language Models: Attacks, Defenses and Evaluationscited by 67
  4. 2025Backdoor Cleaning without External Guidance in MLLM Fine-tuningcited by 35
  5. 2025HiddenDetect: Detecting Jailbreak Attacks against Large Vision-Language Models via Monitoring Hidden Statescited by 33
  6. 2025SafeGRPO: Self-Rewarded Multimodal Safety Alignment via Rule-Governed Policy Optimizationcited by 12
  7. 2025SafePTR: Token-Level Jailbreak Defense in Multimodal LLMs via Prune-then-Restore Mechanismcited by 10
  8. 2025Attention Misses Visual Risk: Risk-Adaptive Steering for Multimodal Safety Alignmentcited by 4
  9. 2025Q-MLLM: Vector Quantization for Robust Multimodal Large Language Model Securitycited by 2
  10. 2025MPCAR: Multi-Perspective Contextual Augmentation for Enhanced Visual Reasoning in Large Vision-Language Modelscited by 1
  11. 2025M3PO: Multimodal-Model-Guided Preference Optimization for Visual Instruction Followingcited by 1
  12. 2025Security Tensors as a Cross-Modal Bridge: Extending Text-Aligned Safety to Vision in LVLMcited by 0
7 more
  1. 2026Safety-Potential Pruning for Enhancing Safety Prompts Against VLM Jailbreaking Without Retrainingcited by 0
  2. 2026Evolving Safety Landscape of Multi-modal Large Language Models: A Survey of Emerging Threats and Safeguardscited by 0
  3. 2026The Safeguard Worked. Is the LLM System Safer?cited by 0
  4. 2026SPARK: Representation-Level KV Memory Alignment for Safer Vision-Language Modelscited by 0
  5. 2026SafeMol: Dual-Modality Safety Alignment for Molecular Multimodal Modelscited by 0
  6. 2026Can Vision-Language Models Stay Helpful When Facing Implicit Risks? Intent-Privilege OPSD for Efficient Safety-Helpfulness Alignmentcited by 0
  7. 2026Scenes as an Attack Surface: Uncovering Scene-Grounded Safety Failures in MLLMscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Unraveling and Mitigating Safety Alignment Degradation of Vision-Language Models
  2. 2025VLM-Guard: Safeguarding Vision-Language Models via Fulfilling Safety Alignment Gap
  3. 2024Safety Alignment for Vision Language Models
  4. 2024ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference Time
  5. 2025VisuoAlign: Safety Alignment of LVLMs with Multimodal Tree Search
  6. 2024PSA-VLM: Enhancing Vision-Language Model Safety through Progressive Concept-Bottleneck-Driven Alignment
  7. 2026SafeRI: Recognition and Intervention for Token-Level Safety Intervention in Large Vision Language Models
  8. 2026Lingua-SafetyBench: A Benchmark for Safety Evaluation of Multilingual Vision-Language Models