EVA-Bench Data 2.0: 3 Domains, 121 Tools, 213 Scenarios

ServiceNow released an expanded evaluation benchmark for voice agents spanning three enterprise domains: airline customer service, IT support, and healthcare HR. The dataset comprises 213 evaluation scenarios across 121 tools, a significant expansion from the original release. The benchmark underwent rigorous validation against leading AI models to ensure scenarios are both challenging and solvable, while incorporating multilingual support and multiple scenario types including adversarial cases.