iiRecord
Agentic AI Atlas · GSM8K
benchmark:gsm8ka5c.ai
II.
Benchmark JSON

benchmark:gsm8k

Structured · live

GSM8K json

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

File · benchmarks/benchmarks/benchmarks-math.yamlCluster · benchmarks
Record JSON
{
  "id": "benchmark:gsm8k",
  "_kind": "Benchmark",
  "_file": "benchmarks/benchmarks/benchmarks-math.yaml",
  "_cluster": "benchmarks",
  "attributes": {
    "displayName": "GSM8K",
    "homepageUrl": "https://github.com/openai/grade-school-math",
    "kind": "math",
    "targetsKind": "ModelVersion",
    "description": "GSM8K (OpenAI) is 8.5K linguistically diverse grade-school math word\nproblems requiring multi-step arithmetic reasoning.\n"
  },
  "outgoingEdges": [
    {
      "from": "benchmark:gsm8k",
      "to": "skill-area:mathematical-reasoning",
      "kind": "covers",
      "attributes": {
        "attributes": {
          "coverage": "full",
          "weight": 0.1
        }
      }
    }
  ],
  "incomingEdges": [
    {
      "from": "eval-result:gsm8k.gemma-2-27b.001",
      "to": "benchmark:gsm8k",
      "kind": "scored_against",
      "attributes": {}
    },
    {
      "from": "eval-result:gsm8k.claude-sonnet-4-5.001",
      "to": "benchmark:gsm8k",
      "kind": "scored_against",
      "attributes": {}
    },
    {
      "from": "eval-run:gsm8k.gemma-2-27b.2024-06",
      "to": "benchmark:gsm8k",
      "kind": "for_benchmark"
    },
    {
      "from": "eval-run:gsm8k.claude-sonnet-4-5.2025-09",
      "to": "benchmark:gsm8k",
      "kind": "for_benchmark"
    },
    {
      "from": "test-set:gsm8k-test",
      "to": "benchmark:gsm8k",
      "kind": "belongs_to_benchmark"
    },
    {
      "from": "scope-boundary:gsm8k.scope",
      "to": "benchmark:gsm8k",
      "kind": "bounds_subject"
    }
  ]
}