mlsanityData to 14 Oct 2026

Can't See the Forest for the Trees: Benchmarking Multimodal Safety Awareness for Multimodal LLMs

arXiv 2502.11184 ↗First posted 16 Feb 2025cs.CL cs.AI cs.CVcited by 23

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

Abstract (arXiv)

Multimodal Large Language Models (MLLMs) have expanded the capabilities of traditional language models by enabling interaction through both text and images. However, ensuring the safety of these models remains a significant challenge, particularly in accurately identifying whether multimodal content is safe or unsafe-a capability we term safety awareness. In this paper, we introduce MMSafeAware, the first comprehensive multimodal safety awareness benchmark designed to evaluate MLLMs across 29 safety scenarios with 1500 carefully curated image-prompt pairs. MMSafeAware includes both unsafe and over-safety subsets to assess models abilities to correctly identify unsafe content and avoid over-sensitivity that can hinder helpfulness. Evaluating nine widely used MLLMs using MMSafeAware reveals that current models are not sufficiently safe and often overly sensitive; for example, GPT-4V misclassifies 36.1% of unsafe inputs as safe and 59.9% of benign inputs as unsafe. We further explore three methods to improve safety awareness-prompting-based approaches, visual contrastive decoding, and vision-centric reasoning fine-tuning-but find that none achieve satisfactory performance. Our findings highlight the profound challenges in developing MLLMs with robust safety awareness, underscoring the need for further research in this area. All the code and data will be publicly available to facilitate future research.

Key facts. Cited by 23 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 17 of its 38 references that are in the map. First posted on arXiv 16 Feb 2025.

Builds on

17 of 38 references.

Closest ancestors · 2

Same direction, newest first

  1. 2024RapGuard: Safeguarding Multimodal Large Language Models via Rationale-aware Defensive Prompting25 Dec
  2. 2024MOSSBench: Is Your Multimodal Language Model Oversensitive to Safe Queries?22 Jun

Other references · 15

  1. 2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,720
  2. 2023BeaverTails: Towards Improved Safety Alignment of LLM via a Human-Preference Datasetcited by 860
  3. 2017Automated Hate Speech Detection and the Problem of Offensive Languagecited by 736
  4. 2023Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decodingcited by 688
  5. 2020The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memescited by 490
  6. 2024LLaVA-CoT: Let Vision Language Models Reason Step-by-Stepcited by 454
  7. 2022ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detectioncited by 422
  8. 2021BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generationcited by 412
  9. 2022SafeText: A Benchmark for Exploring Physical Safety in Language Modelscited by 41
  10. 2023Multimodal ChatGPT for Medical Applications: an Experimental Study of GPT-4Vcited by 38
  11. 2024Automatically Generating UI Code from Screenshot: A Divide-and-Conquer-Based Approachcited by 24
  12. 2024MRWeb: An Exploration of Generating Multi-Page Resource-Aware Web Code from UI Designscited by 24
3 more
  1. 2024New Job, New Gender? Measuring the Social Bias in Image Generation Modelscited by 16
  2. 2024Chain-of-Jailbreak Attack for Image Generation Models via Editing Step by Stepcited by 11
  3. 2024FairGen: Enhancing Fairness in Text-to-Image Diffusion Models via Self-Discovering Latent Directionscited by 8

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 10; 2026: 13.

23 papers.

Sort:

By area:

  1. 2025USB: A Comprehensive and Unified Safety Evaluation Benchmark for Multimodal Large Language Modelscited by 10
  2. 2025SaFeR-VLM: Toward Safety-aware Fine-grained Reasoning in Multimodal Modelscited by 10
  3. 2025Multimodal Cultural Safety: Evaluation Framework and Alignment Strategiescited by 8
  4. 2025Tiered Agentic Oversight: A Hierarchical Multi-Agent System for Healthcare Safetycited by 8
  5. 2025EVADE-Bench: Multimodal Benchmark for Evaluating and Enhancing Evasive Content Detectioncited by 5
  6. 2026Evolving Contextual Safety in Multi-Modal Large Language Models via Inference-Time Self-Reflective Memorycited by 5
  7. 2026MTMCS-Bench: Evaluating Contextual Safety of Multimodal Large Language Models in Multi-Turn Dialoguescited by 4
  8. 2025Protect: Towards Robust Guardrailing Stack for Trustworthy Enterprise LLM Systemscited by 3
  9. 2026When Helpers Become Hazards: A Benchmark for Analyzing Multimodal LLM-Powered Safety in Daily Lifecited by 3
  10. 2026On the Adversarial Robustness of Multimodal LLM Judgescited by 3
  11. 2025Falcon: A Cross-Modal Evaluation Dataset for Comprehensive Safety Perceptioncited by 2
  12. 2026Lingua-SafetyBench: A Benchmark for Safety Evaluation of Multilingual Vision-Language Modelscited by 2
11 more
  1. 2026CSR-Bench: A Benchmark for Evaluating the Cross-modal Safety and Reliability of MLLMscited by 2
  2. 2025Correspondence of high-dimensional emotion structures elicited by video clips between humans and Multimodal LLMscited by 1
  3. 2025LLaVAShield: Safeguarding Multimodal Multi-Turn Dialogues in Vision-Language Modelscited by 1
  4. 2026Toward Accountable AI-Generated Content on Social Platforms: Steganographic Attribution and Multimodal Harm Detectioncited by 1
  5. 2026SafeSteer: A Decoding-level Defense Mechanism for Multimodal Large Language Modelscited by 1
  6. 2026MCBench: A Multicontext Safety Assessment Benchmark for Omni Large Language Modelscited by 1
  7. 2025STEMTOX: From Collaborative Tags to Fine-Grained Toxic Meme Detection via Entropy-Guided Multi-Task Learningcited by 0
  8. 2026HarmVideoBench: Benchmarking Harmful Video Understanding in Large Multimodal Modelscited by 0
  9. 2026SafeGesture: Evaluating Fine-Grained Hand Gesture Understanding in Vision-Language Models through Scenario-Conditioned Safety Interpretationcited by 0
  10. 2026ReFrame: Evidence-Guided Test-Time Safety Alignment in Multimodal Large Language Modelscited by 0
  11. 2026MLLMs Fail to Refuse when Using Tools Agenticallycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Multimodal Situational Safety
  2. 2026Transfer Safety Awareness for Cross-Modal Safety Drift in Multimodal Large Language Models
  3. 2024CoCA: Regaining Safety-awareness of Multimodal Large Language Models with Constitutional Calibration
  4. 2025Steering Multimodal Large Language Models Decoding for Context-Aware Safety
  5. 2024Eyes Closed, Safety On: Protecting Multimodal LLMs via Image-to-Text Transformation
  6. 2025MSTS: A Multimodal Safety Test Suite for Vision-Language Models
  7. 2026MME-Safety: A Fine-grained Benchmark for Safety Evaluation of MLLMs
  8. 2024VLSBench: Unveiling Visual Leakage in Multimodal Safety