List your company
ResearchFormerly WizWand

Scam Detection Benchmark Single-App Content

99.7AccuracyORACLE
93.87695.38896.998.412May 9, 2026
Updated 5mo ago

Evaluation results

MethodLinks
ORACLEBackbone=Qwen3-4B, Category=ORACLE2026.0599.7
FraudR1Category=Specialized Methods2026.0598.9
SafeQAQCategory=Specialized Methods2026.0598.4
ORACLEBackbone=Qwen3-1.7B, Category=ORACLE2026.0598.1
ScamGPT-JCategory=Specialized Methods2026.0597.8
AntiFraudCategory=Specialized Methods2026.0597.5
Claude-4.1-OpusCategory=General LLMs2026.0596.8
Claude-4-SonnetCategory=General LLMs2026.0596.4
GPT-5.1Category=General LLMs2026.0596.3
ORACLEBackbone=Qwen3-0.6B, Category=ORACLE2026.0596.2
BotWarCategory=Specialized Methods2026.0596.1
Gemini-3-FlashCategory=General LLMs2026.0595.9
Kimi-2.5Category=General LLMs2026.0595.7
GLM-4.6Category=General LLMs2026.0595.6
Grok-4Category=General LLMs2026.0594.8
DeepSeek-V4Category=General LLMs2026.0594.1