M3 CHEM Cross-Platform Multimodal Hate Speech Dataset

A 4chan, Twitter, and Weibo (cross-platform) dataset of 2,455 multimodal posts (1,318 hate, 1,137 non-hate), collected by Mira AI Lab (2024). Each record contains image filename, image OCR text, post text, hate/non-hate binary label, hate category (gender/race/religion/politics/violence/health/etc), and human-annotated reason for classification. Multimodal cross-platform collection hate speech dataset combining image OCR text and post text with fine-grained category labels across 8 hate dimensions — 54% of posts are hate speech with human-annotated reasoning. Suitable for multimodal hate speech detection, cross-platform analysis, hate category classification, image-text hate speech, content moderation research. The package includes the 2,455 meme images (img/) that the records name.

#Cross platform#Hate category#M3#Mira ai lab#Multimodal#Ocr

Download data BibTeX JSONSimilar datasets

Analytics

Content metrics

2,455 records scored

is hate53.7% true
true 53.7%false 46.3%
cat gender7.0% true
true 7.0%false 93.0%
cat health status4.3% true
true 4.3%false 95.7%
cat international relations7.3% true
true 7.3%false 92.7%
cat politics14.2% true
true 14.2%false 85.8%
cat public health3.5% true
true 3.5%false 96.5%
cat race13.6% true
true 13.6%false 86.4%
cat religion4.4% true
true 4.4%false 95.6%
cat violence12.1% true
true 12.1%false 87.9%
6.2
Flesch–Kincaid grade
US school grade level
12.3
Automated readability index
Frequent terms
peoplelikedonwhitefuckinggoodjewstime000fuckwomennewtrumpknowrightmanshit组图共2张2024niggerque组图共9张wantjewishgoing组图共3张americarussiathinkukraineworlddayniggersjewlooklivegodisraellovecountrywarracestop组图共4张faggotkikerealneedpostyears
Topic clusters

Topic 0

trump
biden
shill
election
vote
america
joe
maga

Topic 1

russia
ukraine
russian
war
lose
ukrainian
putin
russians

Topic 2

nigger
white
people
like
jews
black
race
fuck

Topic 3

jews
jewish
jew
christ
jesus
king
christians
hate

Topic 4

woman
people
man
autistic
immune
gay
propaganda
autism

Topic 5

میں
follow
pakistan
نہیں
ہیں
السلام
khan
butt

Topic 6

peso
去你妈的日本鬼子
mejor
menor
mayor
工作室
罗云熙
星期一

Topic 7

jest
por
con
para
los
las
del
não

Topic 8

把这给脸不要脸
的畜生打入天牢
register
health
heart
source
good
ang

Topic 9

prabowo
ini
الله
orang
bir
siapa
genocide
dia

Topic 10

leader
practice
infographic
央视新闻
leadership
intelligence
digital
network

Topic 11

bwc
touch
man
aids
ccp
powerful
liu
translate
Show all 18 topics

Topic 12

mode
update
map
base
iss
poccna
rea
novax

Topic 13

jan
usd
cash
user
ice
nft
account
max

Topic 14

happy
thank
follower
week
day
bring
good
friday

Topic 15

vax
cancer
covid
die
unvaccinate
protein
spike
vaxx

Topic 16

retarded
fat
retard
stupid
kid
juice
extra
idiot

Topic 17

india
field
busy
public
road
normal
pooplska
shit
Top hashtags
#任嘉伦17
#每日一善13
#阳光信用12
#早安11
#罗云熙8
#白鹿7
#15
#任嘉伦烈焰之武庚纪5
#巴以冲突5
#pakvsnz4
#普法课堂4
#王鹤棣4
#翔霖4
#转发接龙开新年4
#黄媒不会说的真相4
#bitcoin3
#bl3
#cy3
#digitaltransformation3
#freepalestine3
Top emoji
🤔30
💎29
👇27
🔥26
😂25
🏻21
🙏17
🇸15
𝓉13
👀13
𝓮12
🎉12
🔶12
🤣12
𝒶11

Computed over the full corpus.

Record

Catalog key
m3-chem
Authors
Xing, Rui; Chai, Qi; Ma, Jie; Tao, Jing; Wang, Pinghui; Zhang, Shuming; Wang, Xinping; Wang, Hao
Year
2026
Source paper
Is AI Ready for Multimodal Hate Speech Detection? A Comprehensive Dataset and Benchmark Evaluation
DOI / arXiv
10.48550/arXiv.2603.21686
License
Academic research use — see repository (LicenseRef-Research-Only)
Records
2,455
Nodes are
unknown
Edges are
unknown
Text corpus
social media post text and image OCR text
Languages
en, zh
Collected
2024 (2020s)
Tasks
Hate speech detection (binary; is_hate) inferred
Region
Global (inferred: the platform is used worldwide)

Provenance

Originally published at
https://github.com/mira-ai-lab/M3 link checked
Hosted copy
Mirrored on echo.directory on 2026-10-21, 248.7 MB across 2463 files. Checksums in MANIFEST.json.

Citations

Is AI Ready for Multimodal Hate Speech Detection? A Comprehensive Dataset and Benchmark Evaluation
Rui Xing, Qi Chai, Jie Ma, Jing Tao, Ping-Hui Wang
Journal Article · arXiv.org · 2026

2
citations
few citations so far
0
influential citations
0% of all citations

Cited by

Citation data from Semantic Scholar, retrieved 2026-10-22.

How to cite

BibTeX

@misc{echo2026,
  title  = {ECHO: Exploratory Compendium of Humanity Online},
  author = {Switzer, Braden and Zafarani, Reza},
  year   = {2026},
  url    = {https://echo.directory},
  note   = {Syracuse University}
}

@misc{https://doi.org/10.48550/arxiv.2603.21686,
  doi       = {10.48550/ARXIV.2603.21686},
  url       = {https://arxiv.org/abs/2603.21686},
  author    = {Xing, Rui and Chai, Qi and Ma, Jie and Tao, Jing and Wang, Pinghui and Zhang, Shuming and Wang, Xinping and Wang, Hao},
  keywords  = {Multiagent Systems (cs.MA), FOS: Computer and information sciences, FOS: Computer and information sciences},
  title     = {Is AI Ready for Multimodal Hate Speech Detection? A Comprehensive Dataset and Benchmark Evaluation},
  publisher = {arXiv},
  year      = {2026},
  copyright = {arXiv.org perpetual, non-exclusive license}
}

APA

Switzer, B., & Zafarani, R. (2026). ECHO: Exploratory Compendium of Humanity Online [Data repository]. Syracuse University. https://echo.directory

Xing, R., Chai, Q., Ma, J., Tao, J., Wang, P., Zhang, S., Wang, X., & Wang, H. (2026). M3 CHEM Cross-Platform Multimodal Hate Speech Dataset [Data set]. https://doi.org/10.48550/arXiv.2603.21686

MLA

Switzer, Braden, and Reza Zafarani. ECHO: Exploratory Compendium of Humanity Online. Syracuse University, 2026, echo.directory.

Xing, Rui, et al. M3 CHEM Cross-Platform Multimodal Hate Speech Dataset. 2026, doi.org/10.48550/arXiv.2603.21686.

BibTeX · RIS · CSL-JSON

Similar datasets 10 all with scores →