ads-mm-benchmark / benchmark_tasks /benchmark_summary.json
Cornada's picture
Upload benchmark_tasks/benchmark_summary.json with huggingface_hub
fff11ba verified
[
{
"name": "video_availability_prediction",
"description": "Predict whether a course includes video lectures from its text description",
"type": "binary_classification",
"metric": "F1 (macro), Accuracy",
"n_total": 1137,
"n_train": 822,
"n_dev": 176,
"n_test": 139,
"positive_rate": 1.0
},
{
"name": "cross_market_occupation_matching",
"description": "Match occupations between O*NET and ESCO systems",
"type": "retrieval / matching",
"metric": "Recall@k, MRR",
"n_total": 3039,
"n_train": 2124,
"n_dev": 423,
"n_test": 492
},
{
"name": "ai_exposure_prediction",
"description": "Predict AI automation exposure score (1-10) from occupation text",
"type": "regression",
"metric": "R², MAE, Spearman rho",
"n_total": 0,
"n_train": 0,
"n_dev": 0,
"n_test": 0,
"label_mean": 0,
"label_std": 0
},
{
"name": "cross_lingual_alignment",
"description": "Align same occupations described in different languages",
"type": "retrieval / bitext mining",
"metric": "Recall@1, Recall@5, MRR",
"n_total": 1500,
"n_train": 1067,
"n_dev": 204,
"n_test": 229,
"lang_pairs": [
"en↔uk",
"sv↔uk",
"en↔sv"
]
},
{
"name": "temporal_drift_prediction",
"description": "Predict future occupation embedding from 3 historical versions",
"type": "regression (embedding prediction)",
"metric": "Cosine similarity, L2 distance to ground truth",
"n_total": 1016,
"n_train": 693,
"n_dev": 176,
"n_test": 147
}
]