Prompt
Given order + policy fixtures, approve or deny refund. Graded pass@3 over three trials with identical inputs; DB final state must match gold each trial.
Success criteria
- Correct approve/deny each trial
- pass@3 = 1.0
- Policy citation in reply
Scoring hints
- τ-bench-style pass^k
- reliabilityConsistency primary
Budget 700s
Tools required · web no · code no
Tags refund, policy, multi-trial, pass@3
Primary type Customer-support agent. Gold is not in this payload. API