II.
TestSet JSON
Structured · livetest-set:mle-bench-competitions
MLE-bench Kaggle competition set json
Inspect the normalized record payload exactly as the atlas UI reads it.
{
"id": "test-set:mle-bench-competitions",
"_kind": "TestSet",
"_file": "benchmarks/test-sets/source-backed-benchmark-test-sets.yaml",
"_cluster": "benchmarks",
"attributes": {
"displayName": "MLE-bench Kaggle competition set",
"benchmarkId": "benchmark:mle-bench",
"caseCount": 75,
"releasedAt": "2024-10-09",
"composition": "Curated set of 75 ML engineering competitions from Kaggle, with\npreparation scripts for train/test splits and competition-style\nsubmission grading.\n",
"homepageUrl": "https://openai.com/index/mle-bench/",
"description": "Canonical MLE-bench artifact for end-to-end machine-learning\nengineering agent evaluation.\n"
},
"outgoingEdges": [
{
"from": "test-set:mle-bench-competitions",
"to": "benchmark:mle-bench",
"kind": "belongs_to_benchmark"
}
],
"incomingEdges": []
}