Auto-Generated Rubric Evaluators: Building Context-Aware Evaluators for AI Agents

Authors: Shuo Qiu, Sydney Lister, Ilya Matiach, Ali Mahmoudzadeh, Salma Elshafey, José Santos, Vivek Bhadauria, Morteza Ziyadi, April Kwong Why Your Agent Needs a Task-Specific Evaluator Picture a customer-service agent for a telecom company. A customer messages in asking to switch plans and get a refund for last month’s overcharge. The agent needs to verify the customer’s identity and confirm the new plan before ending the conversation.