II.
TestSet JSON
Structured · livetest-set:assistantbench-214-web-tasks
AssistantBench 214 web tasks json
Inspect the normalized record payload exactly as the atlas UI reads it.
{
"id": "test-set:assistantbench-214-web-tasks",
"_kind": "TestSet",
"_file": "benchmarks/test-sets/source-backed-benchmark-test-sets.yaml",
"_cluster": "benchmarks",
"attributes": {
"displayName": "AssistantBench 214 web tasks",
"benchmarkId": "benchmark:assistant-bench",
"caseCount": 214,
"composition": "AssistantBench's realistic web-agent tasks cover time-consuming\ninformation-seeking and assistant workflows across many websites\nand domains, with automatic evaluation.\n",
"homepageUrl": "https://assistantbench.github.io/",
"description": "Canonical AssistantBench artifact for web-research and web-agent\ntask completion evaluation.\n"
},
"outgoingEdges": [
{
"from": "test-set:assistantbench-214-web-tasks",
"to": "benchmark:assistant-bench",
"kind": "belongs_to_benchmark"
}
],
"incomingEdges": []
}