II.
Benchmark overview
Reference · livebenchmark:assistant-bench
AssistantBench overview
Real-world web-research tasks (price comparison, scheduling, finding multi-attribute info) constructed to stress-test agentic assistants beyond simple QA.
Attributes
displayName
AssistantBench
homepageUrl
kind
web-agent
targetsKind
AgentVersion
description
Real-world web-research tasks (price comparison, scheduling,
finding multi-attribute info) constructed to stress-test agentic
assistants beyond simple QA.
Outgoing edges
applies_to1
- domain:web-development·DomainWeb Development
covers2
- skill-area:deep-web-research·SkillAreaDeep Web Research
- skill-area:agentic-loops·SkillAreaAgentic Loops
Incoming edges
belongs_to_benchmark1
- test-set:assistantbench-214-web-tasks·TestSetAssistantBench 214 web tasks