List your company

Video Question Answering on VideoAgent Aggregate: EgoSchema and NExT-QA

68.606Overall AccuracyJailAgent
48.409253.652658.89664.1394Apr 7, 2026

Evaluation results

MethodLinks
JailAgentAgent Backbone=Gemini-3.0-pro2026.0468.606
JailAgentAgent Backbone=GPT-52026.0466.346
JailAgentAgent Backbone=GPT-4o2026.0463.849
JailAgentAgent Backbone=Claude-3.5-haiku2026.0463.815
BadChainAgent Backbone=Gemini-3.0-pro2026.0463.193
AgentPoisonAgent Backbone=Gemini-3.0-pro2026.0461.886
JailAgentAgent Backbone=GPT-3.5-turbo2026.0461.366
JailAgentAgent Backbone=Llama-3.1-70B2026.0461.057
AgentPoisonAgent Backbone=GPT-4o2026.0461.034
AgentPoisonAgent Backbone=Claude-3.5-haiku2026.0460.661
AgentPoisonAgent Backbone=GPT-52026.0459.351
BadChainAgent Backbone=GPT-52026.0459.159
BadChainAgent Backbone=GPT-3.5-turbo2026.0459.064
BadChainAgent Backbone=Llama-3.1-70B2026.0457.666
BadChainAgent Backbone=GPT-4o2026.0457.568
AgentPoisonAgent Backbone=Llama-3.1-70B2026.0457.55
AgentPoisonAgent Backbone=GPT-3.5-turbo2026.0457.454
PAIRAgent Backbone=Gemini-3.0-pro2026.0455.839
BadChainAgent Backbone=Claude-3.5-haiku2026.0454.834
PAIRAgent Backbone=GPT-4o2026.0452.648
PAIRAgent Backbone=GPT-52026.0452.573
PAIRAgent Backbone=Claude-3.5-haiku2026.0450.456
PAIRAgent Backbone=Llama-3.1-70B2026.0450.082
PAIRAgent Backbone=GPT-3.5-turbo2026.0449.186