List your company

Red-teaming on HarmBench Claude-Sonnet-3.5 (held-out test)

60ASRAGENTICRED
10.0823.043648.96Jan 20, 2026

Evaluation results

MethodLinks
AGENTICRED2026.0160
CoP2026.0138
AdvReasoning2026.0136
J22026.0135
AutoDAN-Turbo2026.0112