
Students can manipulate #LLMAsAJudge to change #Assessment output. #PromptInjection changed 29% LLM scores & 98% narrative #feedback. Major problem for #AISecurity #LLMSecurity #GenAISecurity #AiinEducation #TrustworthyAI #Jailbreaking #AdversarialAI medrxiv.org/content/10.64898…
