mlsanityData to 14 Oct 2026

AI vs. Human Moderators: A Comparative Evaluation of Multimodal LLMs in Content Moderation for Brand Safety

arXiv 2508.05527 ↗First posted 7 Aug 2025cs.CVcited by 8

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

Abstract (arXiv)

As the volume of video content online grows exponentially, the demand for moderation of unsafe videos has surpassed human capabilities, posing both operational and mental health challenges. While recent studies demonstrated the merits of Multimodal Large Language Models (MLLMs) in various video understanding tasks, their application to multimodal content moderation, a domain that requires nuanced understanding of both visual and textual cues, remains relatively underexplored. In this work, we benchmark the capabilities of MLLMs in brand safety classification, a critical subset of content moderation for safe-guarding advertising integrity. To this end, we introduce a novel, multimodal and multilingual dataset, meticulously labeled by professional reviewers in a multitude of risk categories. Through a detailed comparative analysis, we demonstrate the effectiveness of MLLMs such as Gemini, GPT, and Llama in multimodal brand safety, and evaluate their accuracy and cost efficiency compared to professional human reviewers. Furthermore, we present an in-depth discussion shedding light on limitations of MLLMs and failure cases. We are releasing our dataset alongside this paper to facilitate future research on effective and responsible brand safety and content moderation.

Key facts. Cited by 8 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 10 of its 23 references that are in the map. First posted on arXiv 7 Aug 2025.

Builds on

10 of 23 references.

References · 10

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  3. 2022Robust Speech Recognition via Large-Scale Weak Supervisioncited by 4,057
  4. 2020Exploiting Cloze Questions for Few Shot Text Classification and Natural Language Inferencecited by 938
  5. 2017Automated Hate Speech Detection and the Problem of Offensive Languagecited by 736
  6. 2016YouTube-8M: A Large-Scale Video Classification Benchmarkcited by 324
  7. 2024Large Language Models for Data Annotation and Synthesis: A Surveycited by 205
  8. 2023AnnoLLM: Making Large Language Models to Be Better Crowdsourced Annotatorscited by 172
  9. 2024Content Moderation by LLM: From Accuracy to Legitimacycited by 30
  10. 2023Beyond Toxic: Toxicity Detection Datasets are Not Enough for Brand Safetycited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 8.

8 papers.

  1. 2026LLM Performance Predictors: Learning When to Escalate in Hybrid Human-AI Moderation Systemscited by 4
  2. 2026AI as Entertainmentcited by 2
  3. 2026Measuring the Prevalence of Policy Violating Content with ML Assisted Sampling and LLM Labelingcited by 1
  4. 2026Not All Attention Heads Contribute to Critical Visual Token Selection: Head-Aware Pruning Matters Morecited by 1
  5. 2026Rethinking Ground Truth: A Case Study on Human Label Variation in MLLM Benchmarkingcited by 0
  6. 2026SafeLens: Deliberate and Efficient Video Guardrails with Fast-and-Slow Screeningcited by 0
  7. 2026UNIVID: Unified Vision-Language Model for Video Moderationcited by 0
  8. 2026Automatic Hard Example Synthesis with Multi-Level Agentic Data Curationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Old Tricks, New Models: How Simple Image Transformations Break Modern AI-based Content Moderation
  2. 2026When Helpers Become Hazards: A Benchmark for Analyzing Multimodal LLM-Powered Safety in Daily Life
  3. 2026Tool-MCoT: Tool Augmented Multimodal Chain-of-Thought for Content Safety Moderation
  4. 2025Towards Safer Social Media Platforms: Scalable and Performant Few-Shot Harmful Content Moderation Using Large Language Models
  5. 2026No One Model Catches Every Harm: Benchmarking Content Moderation Across Safety Scenarios
  6. 2024MLLMGuard: A Multi-dimensional Safety Evaluation Suite for Multimodal Large Language Models
  7. 2025Can't See the Forest for the Trees: Benchmarking Multimodal Safety Awareness for Multimodal LLMs
  8. 2025Filter-And-Refine: A MLLM Based Cascade System for Industrial-Scale Video Content Moderation