COGNITION: From Evaluation to Defense against Multimodal LLM CAPTCHA Solvers

This article likely discusses the development and evaluation of methods to assess and protect against Large Language Models (LLMs) that can solve CAPTCHAs, which are designed to distinguish humans from bots. The focus is on multimodal LLMs, meaning those that process multiple types of data (e.g., text and images). The research likely explores the vulnerabilities of CAPTCHAs and proposes defense mechanisms.