Agentic AI Atlas

II.

Benchmark JSON

benchmark:mt-bench

Structured · live

MT-Bench json

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

File · benchmarks/benchmarks/benchmarks-reasoning.yamlCluster · benchmarks

Record JSON

{
  "id": "benchmark:mt-bench",
  "_kind": "Benchmark",
  "_file": "benchmarks/benchmarks/benchmarks-reasoning.yaml",
  "_cluster": "benchmarks",
  "attributes": {
    "displayName": "MT-Bench",
    "homepageUrl": "https://lmsys.org/blog/2023-06-22-leaderboard/",
    "kind": "reasoning",
    "targetsKind": "ModelVersion",
    "description": "MT-Bench (LMSYS) is a multi-turn open-ended question set graded by\na strong-LLM judge to evaluate conversational quality across writing,\nreasoning, math, coding, extraction, STEM, and humanities.\n"
  },
  "outgoingEdges": [
    {
      "from": "benchmark:mt-bench",
      "to": "skill-area:general-knowledge-reasoning",
      "kind": "covers",
      "attributes": {}
    },
    {
      "from": "benchmark:mt-bench",
      "to": "domain:computer-science",
      "kind": "applies_to",
      "attributes": {}
    }
  ],
  "incomingEdges": []
}