PatronusAI/glider

PatronusAI/glider is a 4 billion parameter language model fine-tuned from Microsoft's Phi-3.5-mini-instruct, developed by Patronus AI. This model is specifically designed for general-purpose evaluation, capable of judging texts, conversations, and RAG setups based on user-defined criteria and rubrics. It was trained on a diverse dataset covering over 183 metrics and 685 domains, including finance and medicine, and supports a maximum sequence length of 8192 tokens, with tested support up to 12,000 tokens. Its primary strength lies in providing detailed, explainable evaluations for various AI outputs.