A curated list of LLM/MLLM guardrails, safety benchmarks, guard models, jailbreak attacks, moderation datasets, and evaluation tools.

RevealVLLMSafetyEval is a comprehensive pipeline for evaluating Vision-Language Models (VLMs) on their compliance with harm-related policies. It automates the creation of adversarial multi-turn datasets and the evaluation of model responses, supporting responsible AI development and red-teaming efforts.

VAlign-Robust: A research framework for quantifying and mitigating semantic hallucination drift in Vision-Language Models (VLMs) under sensory degradation and adversarial noise.