mlsanityData to 13 Oct 2026

Automated Safety Benchmarking: A Multi-agent Pipeline for LVLMs

arXiv 2601.19507 ↗First posted 27 Jan 2026cs.CLcited by 1

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

also in 1 more
Abstract (arXiv)

Large vision-language models (LVLMs) exhibit remarkable capabilities in cross-modal tasks but face significant safety challenges, which undermine their reliability in real-world applications. Efforts have been made to build LVLM safety evaluation benchmarks to uncover their vulnerability. However, existing benchmarks are hindered by their labor-intensive construction process, static complexity, and limited discriminative power. Thus, they may fail to keep pace with rapidly evolving models and emerging risks. To address these limitations, we propose VLSafetyBencher, the first automated system for LVLM safety benchmarking. VLSafetyBencher introduces four collaborative agents: Data Preprocessing, Generation, Augmentation, and Selection agents to construct and select high-quality samples. Experiments validates that VLSafetyBencher can construct high-quality safety benchmarks within one week at a minimal cost. The generated benchmark effectively distinguish safety, with a safety rate disparity of 70% between the most and least safe models.

Key facts. Cited by 1 paper in the map (data through 13 Oct 2026). Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 31 of its 54 references that are in the map. First posted on arXiv 27 Jan 2026.

Builds on

31 of 54 references.

Closest ancestors · 6

Same direction, newest first

  1. 2025SDEval: Safety Dynamic Evaluation for Multimodal Large Language Models8 Aug
  2. 2025USB: A Comprehensive and Unified Safety Evaluation Benchmark for Multimodal Large Language Models26 May
  3. 2025A Survey of Safety on Large Vision-Language Models: Attacks, Defenses and Evaluations14 Feb
  4. 2025Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety2 Feb
  5. 2024VLSBench: Unveiling Visual Leakage in Multimodal Safety29 Nov
  6. 2024Llama Guard 3 Vision: Safeguarding Human-AI Image Understanding Conversations15 Nov

Other references · 25

  1. 2025Qwen2.5-VL Technical Reportcited by 5,499
  2. 2024GPT-4o System Cardcited by 4,536
  3. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,714
  4. 2025InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Modelscited by 1,798
  5. 2025Gemma 3 Technical Reportcited by 1,748
  6. 2023Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversationscited by 1,181
  7. 2024DeepSeek-V3 Technical Reportcited by 777
  8. 2024DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understandingcited by 566
  9. 2024From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipelinecited by 491
  10. 2023GPT-4 Is Too Smart To Be Safe: Stealthy Chat with LLMs via Ciphercited by 422
  11. 2024Ovis: Structural Embedding Alignment for Multimodal Large Language Modelcited by 167
  12. 2025A Comprehensive Survey in LLM(-Agent) Full Stack Safety: Data, Training and Deploymentcited by 114
13 more
  1. 2023How Many Unicorns Are in This Image? A Safety Evaluation Benchmark for Vision LLMscited by 108
  2. 2024Does Refusal Training in LLMs Generalize to the Past Tense?cited by 79
  3. 2024SafeBench: A Safety Evaluation Framework for Multimodal Large Language Modelscited by 68
  4. 2025GuardReasoner-VL: Safeguarding VLMs via Reinforced Reasoningcited by 42
  5. 2024Jailbreak Attacks and Defenses against Multimodal Generative Models: A Surveycited by 32
  6. 2023Beyond Static Datasets: A Deep Interaction Approach to LLM Evaluationcited by 24
  7. 2024Dynamic Multimodal Evaluation with Flexible Complexity by Vision-Language Bootstrappingcited by 22
  8. 2024AutoBench-V: Can Large Vision-Language Models Benchmark Themselves?cited by 20
  9. 2025MMDT: Decoding the Trustworthiness and Safety of Multimodal Foundation Modelscited by 14
  10. 2024LLM-as-an-Interviewer: Beyond Static Testing Through Dynamic LLM Evaluationcited by 13
  11. 2024StructEval: Deepen and Broaden Large Language Model Assessment via Structured Evaluationcited by 9
  12. 2024BenchAgents: Multi-Agent Systems for Structured Benchmark Creationcited by 9
  13. 2024Stochastic Monkeys at Play: Random Augmentations Cheaply Break LLM Safety Alignmentcited by 5

Built on by

Cited by year (2026 to 13 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Story2Proposal: A Scaffold for Structured Scientific Paper Writingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Safe Vision-Language Models via Unsafe Weights Manipulation
  2. 2026A Safety Report on GPT-5.2, Gemini 3 Pro, Qwen3-VL, Grok 4.1 Fast, Nano Banana Pro, and Seedream 4.5
  3. 2025VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration
  4. 2024Safety Alignment for Vision Language Models
  5. 2026TSHA: A Benchmark for Visual Language Models in Trustworthy Safety Hazard Assessment Scenarios
  6. 2025AGENTSAFE: Benchmarking the Safety of Embodied Agents on Hazardous Instructions
  7. 2025MSTS: A Multimodal Safety Test Suite for Vision-Language Models
  8. 2025Self-Aware Safety Augmentation: Leveraging Internal Semantic Understanding to Enhance Safety in Vision-Language Models