[EMNLP'25] A Benchmark for Assessing VLM Safety with Real-World Memes
DongGeon Lee
oneonlee
AI & ML interests
Data-centric natural language processing, AI Safety & Alignment
Recent Activity
upvoted a paper about 18 hours ago
Multimodal Safety Evaluation Should Measure Controllability Beyond Classification upvoted a paper about 18 hours ago
Priming Base Models for Agentic Post-Training: An Open Midtraining Recipe upvoted a paper about 18 hours ago
Knowledge Distillation During Mid-Training Favors Reasoning over Factual RecallOrganizations
datasets 7
oneonlee/SIUO
Viewer • Updated • 334 • 28
oneonlee/Meme-Safety-Bench
Viewer • Updated • 50.8k • 236 • 7
oneonlee/aegis2-train
Viewer • Updated • 30k • 13 • 1
oneonlee/wildguardmix-train
Viewer • Updated • 86.8k • 23
oneonlee/Wiki-NFQA
Viewer • Updated • 1.89k • 17 • 1
oneonlee/wiki-kg-prob
Viewer • Updated • 230 • 4
oneonlee/cleansed_emocontext
Viewer • Updated • 38.4k • 243 • 2