iiRecord
Agentic AI Atlas · RepoBench
benchmark:repobencha5c.ai
II.
Benchmark JSON

benchmark:repobench

Structured · live

RepoBench json

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

File · benchmarks/benchmarks/benchmarks-coding.yamlCluster · benchmarks
Record JSON
{
  "id": "benchmark:repobench",
  "_kind": "Benchmark",
  "_file": "benchmarks/benchmarks/benchmarks-coding.yaml",
  "_cluster": "benchmarks",
  "attributes": {
    "displayName": "RepoBench",
    "homepageUrl": "https://github.com/Leolty/repobench",
    "kind": "code-completion",
    "targetsKind": "ModelVersion",
    "description": "RepoBench evaluates repository-level code auto-completion across\nPython and Java. Sub-tasks measure retrieval, prediction, and\nend-to-end completion in real-world multi-file repositories.\n"
  },
  "outgoingEdges": [
    {
      "from": "benchmark:repobench",
      "to": "domain:software-engineering",
      "kind": "applies_to",
      "attributes": {}
    }
  ],
  "incomingEdges": [
    {
      "from": "scope-boundary:repobench.scope",
      "to": "benchmark:repobench",
      "kind": "bounds_subject"
    }
  ]
}