II.
TestSet JSON
Structured · livetest-set:androidworld-programmatic-tasks
AndroidWorld programmatic task suite json
Inspect the normalized record payload exactly as the atlas UI reads it.
{
"id": "test-set:androidworld-programmatic-tasks",
"_kind": "TestSet",
"_file": "benchmarks/test-sets/source-backed-benchmark-test-sets.yaml",
"_cluster": "benchmarks",
"attributes": {
"displayName": "AndroidWorld programmatic task suite",
"benchmarkId": "benchmark:android-world",
"caseCount": 116,
"composition": "AndroidWorld's reproducible mobile-agent task suite covers real\nAndroid apps with programmatic rewards and dynamically-instantiated\nnatural-language tasks.\n",
"homepageUrl": "https://google-research.github.io/android_world/",
"description": "Canonical AndroidWorld artifact for autonomous Android UI-control\nevaluation.\n"
},
"outgoingEdges": [
{
"from": "test-set:androidworld-programmatic-tasks",
"to": "benchmark:android-world",
"kind": "belongs_to_benchmark"
}
],
"incomingEdges": []
}