Benchmark RadarAI BENCHMARK PROFILE
REDAgentBench
Evaluates LLM agent safety through executable red-teaming, adversarial case generation, and verification of harmful effects across 1,661 cases and five service surfaces.
- Released
- 2026-08-11
- Readiness
- Paper only
- Primary field
- Cybersecurity
Why it matters
Provides an executable and measurable approach for agent safety evaluation beyond aggregate attack success rates.
Motivation
Large language model (LLM) agents combine language-based reasoning with external tools to perform complex tasks.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.