#keepAIopen

I'm thrilled to announce LlavaGuard, a family of VLM-based safeguard models that offers a versatile framework for evaluating the safety compliance of visual content. ml-research.github.io/human-centered…

🚨Excited to share that the V2 of the social impact paper, led by the incredible Irene Solaiman and Zeerak (زیرک ظلعت) is on Zeerak@ mastodon|bsky, is finally out! In this work, we present a guide for evaluating the social impact of Gen AI systems across categories & modalities. 🧵(1/7) arxiv.org/pdf/2306.05949

Today, we are releasing MSTS, a new Multimodal Safety Test Suite for vision-language models! MSTS is exciting because it tests for safety risks *created by multimodality*. Each prompt consists of a text + image that *only in combination* reveal their full unsafe meaning. 🧵