Detecting Offensive Memes with Social Biases in Singapore Context Using Multimodal Large Language Models

• URL: http://arxiv.org/abs/2502.18101v2
• Date: Sat, 08 Mar 2025 08:35:02 GMT
• Title: Detecting Offensive Memes with Social Biases in Singapore Context Using Multimodal Large Language Models
• Abstract summary: Traditional online content moderation systems struggle to classify modern multimodal means of communication, such as memes.<n>We curate a large collection of 112K memes labeled by GPT-4V for fine-tuning a VLM to classify offensive memes in Singapore context.<n>Our solutions reach 80.62% accuracy and 0.8192 AUROC on a held-out test set, and can greatly aid human in moderating online contents.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Traditional online content moderation systems struggle to classify modern multimodal means of communication, such as memes, a highly nuanced and information-dense medium. This task is especially hard in a culturally diverse society like Singapore, where low-resource languages are used and extensive knowledge on local context is needed to interpret online content. We curate a large collection of 112K memes labeled by GPT-4V for fine-tuning a VLM to classify offensive memes in Singapore context. We show the effectiveness of fine-tuned VLMs on our dataset, and propose a pipeline containing OCR, translation and a 7-billion parameter-class VLM. Our solutions reach 80.62% accuracy and 0.8192 AUROC on a held-out test set, and can greatly aid human in moderating online contents. The dataset, code, and model weights have been open-sourced at https://github.com/aliencaocao/vlm-for-memes-aisg.

Related papers

• MemeCULT-1K: Benchmarking South Asian Cultural Context and Humor Understanding of Multimodal Models [7.9152408875072275]
  We introduce MemeCULT-1K, a benchmark of 1,000 South Asian memes in Bengali, English, and Hindi.<n>We evaluate thirteen popular Vision Language Models (VLMs) under two settings: meme-only and context-aware.<n>Fine-grained error analysis reveals that closed-source models fail mainly on entity and reference misidentification, while open-source models are bottlenecked by broader cultural knowledge gaps.
  arXiv  Detail & Related papers  (2026-09-01T18:43:03Z)
• MemeLens: Multilingual Multitask VLMs for Memes [45.8232386994625]
  We propose MemeLens, a unified multilingual and explanation-enhanced Vision Language Model (VLM) for meme understanding.<n>We consolidate 38 public meme datasets, filter and map dataset-specific labels into a shared taxonomy of $20$ tasks spanning harm, targets, figurative/pragmatic intent, and affect.<n>Our findings suggest that robust meme understanding requires multimodal training, exhibits substantial variation across semantic categories, and remains sensitive to over-specialization when models are fine-tuned on individual datasets rather than trained in a unified setting.
  arXiv  Detail & Related papers  (2026-01-18T19:01:03Z)
• Toxicity Red-Teaming: Benchmarking LLM Safety in Singapore's Low-Resource Languages [57.059267233093465]
  Large Language Models (LLMs) have transformed natural language processing, but their safety mechanisms remain under-explored in low-resource, multilingual settings.<n>We introduce textsfSGToxicGuard, a novel dataset and evaluation framework for benchmarking LLM safety in Singapore's diverse linguistic context.<n>We conduct extensive experiments with state-of-the-art multilingual LLMs, and the results uncover critical gaps in their safety guardrails.
  arXiv  Detail & Related papers  (2025-09-18T08:14:34Z)
• MemeReaCon: Probing Contextual Meme Understanding in Large Vision-Language Models [50.2355423914562]
  We introduce MemeReaCon, a novel benchmark designed to evaluate how Large Vision Language Models (LVLMs) understand memes in their original context.<n>We collected memes from five different Reddit communities, keeping each meme's image, the post text, and user comments together.<n>Our tests with leading LVLMs show a clear weakness: models either fail to interpret critical information in the contexts, or overly focus on visual details while overlooking communicative purpose.
  arXiv  Detail & Related papers  (2025-05-23T03:27:23Z)
• Multi3Hate: Multimodal, Multilingual, and Multicultural Hate Speech Detection with Vision-Language Models [11.82100047858478]
  We create the first multimodal and multilingual parallel hate speech dataset, annotated by a multicultural set of annotators, called Multi3Hate.<n>It contains 300 parallel meme samples across 5 languages: English, German, Spanish, Hindi, and Mandarin.<n>We demonstrate that cultural background significantly affects multimodal hate speech annotation in our dataset. The average pairwise agreement among countries is just 74%, significantly lower than that of randomly selected annotator groups.
  arXiv  Detail & Related papers  (2024-11-06T13:06:43Z)
• MURI: High-Quality Instruction Tuning Datasets for Low-Resource Languages via Reverse Instructions [54.08017526771947]
  Multilingual Reverse Instructions (MURI) generates high-quality instruction tuning datasets for low-resource languages. MURI produces instruction-output pairs from existing human-written texts in low-resource languages. Our dataset, MURI-IT, includes more than 2 million instruction-output pairs across 200 languages.
  arXiv  Detail & Related papers  (2024-09-19T17:59:20Z)
• OSPC: Detecting Harmful Memes with Large Language Model as a Catalyst [23.806625629458946]
  This study presents a novel approach to detecting harmful memes, particularly within the multicultural and multilingual context of Singapore. Our methodology integrates image captioning, Optical Character Recognition (OCR), and Large Language Model (LLM) analysis. Our framework achieves top-1 at the public leaderboard of the Online Safety Prize Challenge hosted by AI Singapore.
  arXiv  Detail & Related papers  (2024-06-14T07:28:02Z)
• mOSCAR: A Large-scale Multilingual and Multimodal Document-level Corpus [52.83121058429025]
  We introduce mOSCAR, the first large-scale multilingual and multimodal document corpus crawled from the web. It covers 163 languages, 315M documents, 214B tokens and 1.2B images. It shows a strong boost in few-shot learning performance across various multilingual image-text tasks and benchmarks.
  arXiv  Detail & Related papers  (2024-06-13T00:13:32Z)
• ArMeme: Propagandistic Content in Arabic Memes [9.48177009736915]
  We develop an Arabic memes dataset with manual annotations of propagandistic content. We provide a comprehensive analysis aiming to develop computational tools for their detection.
  arXiv  Detail & Related papers  (2024-06-06T09:56:49Z)
• Exploring the Limits of Zero Shot Vision Language Models for Hate Meme Detection: The Vulnerabilities and their Interpretations [9.970031080934003]
  We study the effectiveness of modern day vision language models (VLMs) in handling intricate tasks such as hate meme detection. We perform thorough prompt engineering and query state-of-the-art VLMs using various prompt types to detect hateful/harmful memes.
  arXiv  Detail & Related papers  (2024-02-19T15:03:04Z)
• NusaWrites: Constructing High-Quality Corpora for Underrepresented and Extremely Low-Resource Languages [54.808217147579036]
  We conduct a case study on Indonesian local languages. We compare the effectiveness of online scraping, human translation, and paragraph writing by native speakers in constructing datasets. Our findings demonstrate that datasets generated through paragraph writing by native speakers exhibit superior quality in terms of lexical diversity and cultural content.
  arXiv  Detail & Related papers  (2023-09-19T14:42:33Z)
• Chain-of-Dictionary Prompting Elicits Translation in Large Language Models [100.47154959254937]
  Large language models (LLMs) have shown surprisingly good performance in multilingual neural machine translation (MNMT) We present a novel method, CoD, which augments LLMs with prior knowledge with the chains of multilingual dictionaries for a subset of input words to elicit translation abilities.
  arXiv  Detail & Related papers  (2023-05-11T05:19:47Z)
• DisinfoMeme: A Multimodal Dataset for Detecting Meme Intentionally Spreading Out Disinformation [72.18912216025029]
  We present DisinfoMeme to help detect disinformation memes. The dataset contains memes mined from Reddit covering three current topics: the COVID-19 pandemic, the Black Lives Matter movement, and veganism/vegetarianism.
  arXiv  Detail & Related papers  (2022-05-25T09:54:59Z)
• bitsa_nlp@LT-EDI-ACL2022: Leveraging Pretrained Language Models for Detecting Homophobia and Transphobia in Social Media Comments [0.9981479937152642]
  We present our system for the LT-EDI shared task on detecting homophobia and transphobia in social media comments. We experiment with a number of monolingual and multilingual transformer based models such as mBERT. We observe their performance on a carefully annotated, real life dataset of YouTube comments in English as well as Tamil.
  arXiv  Detail & Related papers  (2022-03-27T10:15:34Z)
• FBERT: A Neural Transformer for Identifying Offensive Content [67.12838911384024]
  fBERT is a BERT model retrained on SOLID, the largest English offensive language identification corpus available with over $1.4$ million offensive instances. We evaluate fBERT's performance on identifying offensive content on multiple English datasets and we test several thresholds for selecting instances from SOLID. The fBERT model will be made freely available to the community.
  arXiv  Detail & Related papers  (2021-09-10T19:19:26Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.