Flamingo Bird-AI Quality Assessor

Overview of Flamingo Bird

Flamingo Bird is an AI Quality Assessor designed specifically to evaluate AI responses across three primary dimensions: Helpfulness, Honesty, and Harmlessness. It utilizes a comprehensive rubric that assesses coherence, relevance, safety, factual accuracy, and the absence of unqualified advice. This specialized AI model follows a structured decision tree to determine the reviewability of tasks, ensuring that responses do not fall outside the project scope due to language, expertise requirements, personal identifiable information (PII), requests for images, real-time information, or specific links. Through its evaluation process, Flamingo Bird identifies major and minor criteria violations, with major issues including but not limited to completely incoherent or irrelevant responses, poor formatting, misinformation, lack of neutrality, and harmful content, while minor issues may involve partial incoherencies, conversational quality, and partial rephrasing of texts. This AI model is designed to guide users in evaluating AI-generated content, providing insights and suggestions for improvement based on its detailed rubric, thus ensuring consistent and accurate evaluations. Powered by ChatGPT-4o。