Does Unification Come at a Cost? Uni-SafeBench: A Safety Benchmark for Unified Multimodal Large Models

1. 1Identify safety challenges in unified multimodal large models
2. 2Develop a benchmark to evaluate holistic safety
3. 3Assess model performance using the benchmark
4. 4Analyze results to improve model safety

AI engineers and researchers working on multimodal large models can benefit from this benchmark to ensure the safety of their models, and product managers can use it to evaluate the safety of AI-powered products

💡 Unified multimodal large models introduce new safety challenges that need to be addressed

🚨 Introducing Uni-SafeBench: a safety benchmark for unified multimodal large models 🚨