mlsanityData to 14 Oct 2026

ICM-Assistant: Instruction-tuning Multimodal Large Language Models for Rule-based Explainable Image Content Moderation

arXiv 2412.18216 ↗First posted 24 Dec 2024cs.CV cs.CLcited by 11

On the map: Multimodal LLMs / vision-language models → MLLM training and visual instruction tuning

Abstract (arXiv)

Controversial contents largely inundate the Internet, infringing various cultural norms and child protection standards. Traditional Image Content Moderation (ICM) models fall short in producing precise moderation decisions for diverse standards, while recent multimodal large language models (MLLMs), when adopted to general rule-based ICM, often produce classification and explanation results that are inconsistent with human moderators. Aiming at flexible, explainable, and accurate ICM, we design a novel rule-based dataset generation pipeline, decomposing concise human-defined rules and leveraging well-designed multi-stage prompts to enrich short explicit image annotations. Our ICM-Instruct dataset includes detailed moderation explanation and moderation Q-A pairs. Built upon it, we create our ICM-Assistant model in the framework of rule-based ICM, making it readily applicable in real practice. Our ICM-Assistant model demonstrates exceptional performance and flexibility. Specifically, it significantly outperforms existing approaches on various sources, improving both the moderation classification (36.8% on average) and moderation explanation quality (26.6% on average) consistently over existing MLLMs. Code/Data is available at https://github.com/zhaoyuzhi/ICM-Assistant.

Key facts. Cited by 11 papers in the map (data through 14 Oct 2026). Placed in MLLM training and visual instruction tuning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 15 of its 34 references that are in the map. First posted on arXiv 24 Dec 2024.

Builds on

15 of 34 references.

Closest ancestors · 3

Same direction, newest first

  1. 2024InternLM-XComposer2: Mastering Free-form Text-Image Composition and Comprehension in Vision-Language Large Model29 Jan
  2. 2023ShareGPT4V: Improving Large Multi-Modal Models with Better Captions21 Nov
  3. 2023mPLUG-Owl2: Revolutionizing Multi-modal Large Language Model with Modality Collaboration7 Nov

Other references · 12

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  3. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
  4. 2021Zero-Shot Text-to-Image Generationcited by 4,288
  5. 2023LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Daycited by 1,305
  6. 2023GPTScore: Evaluate as You Desirecited by 361
  7. 2022A Holistic Approach to Undesired Content Detection in the Real Worldcited by 356
  8. 2023Q-Instruct: Improving Low-level Visual Abilities for Multi-modality Foundation Modelscited by 152
  9. 2023Adapting Large Language Models for Content Moderation: Pitfalls in Data Engineering and Supervised Fine-tuningcited by 29
  10. 2023Rethinking Multimodal Content Moderation from an Asymmetric Angle with Mixed-modalitycited by 9
  11. 2022Reliable Decision from Multiple Subtasks through Threshold Optimization: Content Moderation in the Wildcited by 7
  12. 2021VideoModerator: A Risk-aware Framework for Multimodal Video Moderation in E-Commercecited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 7; 2026: 4.

11 papers.

  1. 2025AesBiasBench: Evaluating Bias and Alignment in Multimodal Language Models for Personalized Image Aesthetic Assessmentcited by 11
  2. 2025Better Reasoning with Less Data: Enhancing VLMs Through Unified Modality Scoringcited by 6
  3. 2025Embedding-based Retrieval in Multimodal Content Moderationcited by 4
  4. 2025Towards Trustworthy Multimodal Moderation via Policy-Aligned Reasoning and Hierarchical Labelingcited by 4
  5. 2026Making MLLMs Blind: Adversarial Smuggling Attacks in MLLM Content Moderationcited by 4
  6. 2025Aether Weaver: Multimodal Affective Narrative Co-Generation with Dynamic Scene Graphscited by 1
  7. 2025CompAgent: An Agentic Framework for Visual Compliance Verificationcited by 1
  8. 2025When Harmful Content Gets Camouflaged: Unveiling Perception Failure of LVLMs with CamHarmTIcited by 0
  9. 2026CHAIRO: Contextual Hierarchical Analogical Induction and Reasoning Optimization for LLMscited by 0
  10. 2026Auditing Exposure to Harmful Content on TikTok using Multimodal Language Models: A Cross-National, Age-Stratified Studycited by 0
  11. 2026When MLLMs miss what humans read: diagnosing and learning to read squint textcited 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. 2024Moderator: Moderating Text-to-Image Diffusion Models through Fine-grained Context-based Policies
  3. 2026RuleSafe-VL: Evaluating Rule-Conditioned Decision Reasoning in Vision-Language Content Moderation
  4. 2025A High-Quality Dataset and Reliable Evaluation for Interleaved Image-Text Generation
  5. 2023A Review of Vision-Language Models and their Performance on the Hateful Memes Challenge
  6. 2025Filter-And-Refine: A MLLM Based Cascade System for Industrial-Scale Video Content Moderation
  7. 2025Language Integration in Fine-Tuning Multimodal Large Language Models for Image-Based Regression
  8. 2026Explaining is Harder Than Predicting Alone: Evaluating Concept-based Explanations of MLLMs as ICL Visual Classifiers