.svg)
What is AgentHarm Safety Benchmark?
The AgentHarm Safety Benchmark evaluates how effectively AI agents perform complex, multi-step tasks while maintaining safety. It ensures that agents complete tasks without causing harm or breaching safety standards.
Key Features
- Multi-step Task Evaluation
- Agent Safety Assessment
- Task Completion Testing
- Safety Boundary Evaluation
- Harm Prevention Measurement
Use Cases
- Agent Safety Testing
- Multi-step Task Evaluation
- Safety Mechanism Validation
