II.
Benchmark JSON
Structured · livebenchmark:tau-bench
tau-bench json
Inspect the normalized record payload exactly as the atlas UI reads it.
{
"id": "benchmark:tau-bench",
"_kind": "Benchmark",
"_file": "benchmarks/benchmarks/tau-bench.yaml",
"_cluster": "benchmarks",
"attributes": {
"displayName": "tau-bench",
"homepageUrl": "https://github.com/sierra-research/tau-bench",
"kind": "tool-use",
"targetsKind": "AgentVersion",
"description": "Sierra Research's tau-bench — multi-turn agent benchmark in retail and airline customer-service domains.\n"
},
"outgoingEdges": [
{
"from": "benchmark:tau-bench",
"to": "skill-area:multi-turn-tool-use",
"kind": "covers",
"attributes": {
"attributes": {
"coverage": "full",
"weight": 0.7
}
}
},
{
"from": "benchmark:tau-bench",
"to": "skill-area:agentic-loops",
"kind": "covers",
"attributes": {
"attributes": {
"coverage": "partial",
"weight": 0.3
}
}
},
{
"from": "benchmark:tau-bench",
"to": "domain:retail",
"kind": "applies_to",
"attributes": {}
},
{
"from": "benchmark:tau-bench",
"to": "domain:customer-experience",
"kind": "applies_to",
"attributes": {}
}
],
"incomingEdges": [
{
"from": "scope-boundary:tau-bench.scope",
"to": "benchmark:tau-bench",
"kind": "bounds_subject"
}
]
}