Agentic AI Atlas

II.

Benchmark JSON

benchmark:swe-bench

Structured · live

SWE-bench json

Inspect the normalized record payload exactly as the atlas UI reads it.

File · benchmarks/benchmarks/swe-bench.yamlCluster · benchmarks

Record JSON

{
  "id": "benchmark:swe-bench",
  "_kind": "Benchmark",
  "_file": "benchmarks/benchmarks/swe-bench.yaml",
  "_cluster": "benchmarks",
  "attributes": {
    "displayName": "SWE-bench",
    "homepageUrl": "https://www.swebench.com/",
    "kind": "full-stack",
    "targetsKind": "AgentVersion",
    "description": "Real-world software engineering issues from open-source Python repos.\n"
  },
  "outgoingEdges": [
    {
      "from": "benchmark:swe-bench",
      "to": "skill-area:bug-fixing-from-issues",
      "kind": "covers",
      "attributes": {
        "attributes": {
          "coverage": "full",
          "weight": 0.05
        }
      }
    },
    {
      "from": "benchmark:swe-bench",
      "to": "domain:software-engineering",
      "kind": "applies_to",
      "attributes": {}
    }
  ],
  "incomingEdges": [
    {
      "from": "benchmark:swe-bench-verified",
      "to": "benchmark:swe-bench",
      "kind": "refines"
    },
    {
      "from": "test-set:swe-bench-original",
      "to": "benchmark:swe-bench",
      "kind": "belongs_to_benchmark"
    },
    {
      "from": "scope-boundary:swe-bench.scope",
      "to": "benchmark:swe-bench",
      "kind": "bounds_subject"
    }
  ]
}