List your company

Question Answering on HotpotQA (Utility and ASR)

34ASRNo Defense
-1.367.821726.18Apr 1, 2026

Evaluation results

MethodLinks
No DefenseAttack Type=Combined, Target LLM=Qwen-3-4B-Instruct-25072026.0434—
PromptArmorAttack Type=Combined, Target LLM=Qwen-3-4B-Instruct-25072026.0434—
PromptGuardAttack Type=Combined, Target LLM=Qwen-3-4B-Instruct-25072026.0428—
PIGuardAttack Type=Combined, Target LLM=Qwen-3-4B-Instruct-25072026.0425—
No DefenseAttack Type=Direct, Target LLM=Qwen-3-4B-Instruct-25072026.0413—
PromptArmorAttack Type=Direct, Target LLM=Qwen-3-4B-Instruct-25072026.0413—
PromptGuardAttack Type=Direct, Target LLM=Qwen-3-4B-Instruct-25072026.0413—
PIGuardAttack Type=Direct, Target LLM=Qwen-3-4B-Instruct-25072026.049—
DataSentinelAttack Type=Direct, Target LLM=Qwen-3-4B-Instruct-25072026.043—
DataSentinelAttack Type=Combined, Target LLM=Qwen-3-4B-Instruct-25072026.042—
GPT-OSS-SafeguardAttack Type=Direct, Target LLM=Qwen-3-4B-Instruct-25072026.040—
GPT-OSS-SafeguardAttack Type=Combined, Target LLM=Qwen-3-4B-Instruct-25072026.040—
AgentWatcherAttack Type=Direct, Target LLM=Qwen-3-4B-Instruct-25072026.040—
AgentWatcherAttack Type=Combined, Target LLM=Qwen-3-4B-Instruct-25072026.040—
AgentWatcherAttack Type=Clean, Target LLM=Qwen-3-4B-Instruct-25072026.04—57
DataSentinelAttack Type=Clean, Target LLM=Qwen-3-4B-Instruct-25072026.04—14
GPT-OSS-SafeguardAttack Type=Clean, Target LLM=Qwen-3-4B-Instruct-25072026.04—58
No DefenseAttack Type=Clean, Target LLM=Qwen-3-4B-Instruct-25072026.04—58
PIGuardAttack Type=Clean, Target LLM=Qwen-3-4B-Instruct-25072026.04—56
PromptArmorAttack Type=Clean, Target LLM=Qwen-3-4B-Instruct-25072026.04—58
PromptGuardAttack Type=Clean, Target LLM=Qwen-3-4B-Instruct-25072026.04—58