List your company

ResearchFormerly WizWand

LLM Agent Utility

Benchmarks

DatasetSOTA methodMetricTrendResultsLast updated

AgentHarm Benign Requests

Baseline (No Defense)

69.1Utility Score

23

May 8, 2026

Frontier Models Evaluation Set Benign Queries

1.1Benign Refusal Rate

6

May 8, 2026