List your company

Question Answering on StrategyQA, MMLU, and HotpotQA Combined

0.6334Overall AccuracyJailAgent
0.278260.3704630.4626650.554867Apr 7, 2026

Evaluation results

MethodLinks
JailAgentAgent Backbone=Gemini-3.0-pro2026.040.6334
JailAgentAgent Backbone=GPT-52026.040.6308
JailAgentAgent Backbone=GPT-4o2026.040.6174
JailAgentAgent Backbone=Claude-3.5-haiku2026.040.5846
AgentPoisonAgent Backbone=Gemini-3.0-pro2026.040.5835
AgentPoisonAgent Backbone=GPT-4o2026.040.5828
JailAgentAgent Backbone=GPT-3.5-turbo2026.040.5718
BadChainAgent Backbone=Gemini-3.0-pro2026.040.5706
BadChainAgent Backbone=GPT-4o2026.040.557
AgentPoisonAgent Backbone=GPT-52026.040.5537
AgentPoisonAgent Backbone=Claude-3.5-haiku2026.040.5393
BadChainAgent Backbone=GPT-52026.040.5281
AgentPoisonAgent Backbone=GPT-3.5-turbo2026.040.5256
BadChainAgent Backbone=Claude-3.5-haiku2026.040.5103
JailAgentAgent Backbone=Llama-3.1-70B2026.040.4955
BadChainAgent Backbone=GPT-3.5-turbo2026.040.485
PAIRAgent Backbone=Gemini-3.0-pro2026.040.4755
PAIRAgent Backbone=GPT-52026.040.4728
AgentPoisonAgent Backbone=Llama-3.1-70B2026.040.4651
PAIRAgent Backbone=GPT-4o2026.040.4576
BadChainAgent Backbone=Llama-3.1-70B2026.040.4425
PAIRAgent Backbone=Claude-3.5-haiku2026.040.442
PAIRAgent Backbone=GPT-3.5-turbo2026.040.4343
JailAgentAgent Backbone=ERNIE-3.52026.040.4066
PAIRAgent Backbone=Llama-3.1-70B2026.040.3836
AgentPoisonAgent Backbone=ERNIE-3.52026.040.3544
BadChainAgent Backbone=ERNIE-3.52026.040.3475
PAIRAgent Backbone=ERNIE-3.52026.040.2919