mlsanityData to 14 Oct 2026

One Anchor for All: Unified Multilingual and Multimodal Safety Alignment for LVLMs

arXiv 2607.27917 ↗First posted 30 Jul 2026cs.AIcited by 0

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

also in 2 more
Abstract (arXiv)

As large vision-language models (LVLMs) are deployed globally, the combination of multilingual instructions and visual information makes malicious attacks more covert and sophisticated than ever before. However, existing methods isolate language and modality defenses, which, coupled with the scarcity of safety data and high fine-tuning costs, makes it difficult for models to defend against compound attacks. To address this severe challenge, we propose a neuron-level cross-dimensional safety alignment framework driven by modality- and language-shared safety neurons (MLS-Neurons). First, we identify monolingual and unimodal safety neurons by comparing responses to harmful and benign samples, quantifying functional saliency through activation strength and downstream impact. Then, by intersecting these unimodal neurons within each language, we extract modality-shared safety neurons (MS-Neurons) responsive to both visual and textual risks, bridging the safety representation gap between modalities. Furthermore, using English as a semantic anchor, we intersect MS-Neurons across languages to identify modality- and language-shared safety neurons (MLS-Neurons), serving as key defenses against compound attacks. Finally, we update only this minimal subset of shared neurons (~0.03% of parameters), transferring English-only safety supervision to multilingual and multimodal scenarios. Extensive experiments show that our method significantly outperforms state-of-the-art approaches across diverse multilingual and multimodal safety benchmarks while preserving general utility.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 23 of its 46 references that are in the map. First posted on arXiv 30 Jul 2026.

Builds on

23 of 46 references.

Closest ancestors · 3

Same direction, newest first

  1. 2026SafeNeuron: Neuron-Level Safety Alignment for Large Language Models12 Feb
  2. 2026Lingua-SafetyBench: A Benchmark for Safety Evaluation of Multilingual Vision-Language Models30 Jan
  3. 2024ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference Time9 Oct

Other references · 20

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2025Qwen3 Technical Reportcited by 8,820
  3. 2025Gemma 3 Technical Reportcited by 1,824
  4. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilitiescited by 1,239
  5. 2022Language Models are Multilingual Chain-of-Thought Reasonerscited by 530
  6. 2025Open Problems in Mechanistic Interpretabilitycited by 179
  7. 2024Safety Fine-Tuning at (Almost) No Cost: A Baseline for Vision Large Language Modelscited by 167
  8. 2025Qwen3Guard Technical Reportcited by 155
  9. 2024Language-Specific Neurons: The Key to Multilingual Capabilities in Large Language Modelscited by 126
  10. 2024The Language Barrier: Dissecting Safety Challenges of LLMs in Multilingual Contextscited by 94
  11. 2025Commercial LLM Agents Are Already Vulnerable to Simple Yet Dangerous Attackscited by 49
  12. 2025CultureGuard: Towards Culturally-Aware Dataset and Guard Model for Multilingual Safety Applicationscited by 27
8 more
  1. 2025H-Neurons: On the Existence, Impact, and Origin of Hallucination-Associated Neurons in LLMscited by 22
  2. 2025The Emergence of Abstract Thought in Large Language Models Beyond Any Languagecited by 21
  3. 2026DNA: Uncovering Universal Latent Forgery Knowledgecited by 14
  4. 2026Align Once, Benefit Multilingually: Enforcing Multilingual Consistency for LLM Safety Alignmentcited by 9
  5. 2026BalDRO: A Distributionally Robust Optimization based Framework for Large Language Model Unlearningcited by 8
  6. 2026Who Transfers Safety? Identifying and Targeting Cross-Lingual Shared Safety Neuronscited by 8
  7. 2026Multilingual Safety Alignment Via Sparse Weight Editingcited by 6
  8. 2025Multi-Level Safety Continual Projection for Fine-Tuned Large Language Models without Retrainingcited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Targeted Interpretable Safety Neuron Enhancement for Multilingual Vision-Language Large Models
  2. 2026SafeNexus: Discovering and Steering Modality-Universal Safety Neurons in MLLMs
  3. 2024Cross-Modal Safety Alignment: Is textual unlearning all you need?
  4. 2026Do VLMs Share Safety Neurons Across Modalities?
  5. 2025Security Tensors as a Cross-Modal Bridge: Extending Text-Aligned Safety to Vision in LVLM
  6. 2025Do We Really Need Curated Malicious Data for Safety Alignment in Multi-modal Large Language Models?
  7. 2026Diagnosing and Repairing Unsafe Channels in Vision-Language Models via Causal Discovery and Dual-Modal Safety Subspace Projection
  8. 2024Cross-Modal Safety Mechanism Transfer in Large Vision-Language Models