id stringlengths 16 26 | prompt stringlengths 101 184 | task stringclasses 6
values | domain stringlengths 2 16 | input_schema dict | dataset_profile dict | expected_blueprint dict | rubric dict | failure_modes listlengths 3 3 | source dict |
|---|---|---|---|---|---|---|---|---|---|
fraud_scoring_001 | Build a fraud scoring API. The table has transaction_id, amount, merchant_risk, prior_chargebacks, and an is_fraud label. | classification | fintech | {
"filename": "fraud_scoring_001.csv",
"row_count": 4,
"column_count": 5,
"columns": [
{
"name": "transaction_id",
"kind": "id",
"missing_ratio": 0
},
{
"name": "amount",
"kind": "numeric",
"missing_ratio": 0
},
{
"name": "merchant_risk",
"kind... | {
"filename": "fraud_scoring_001.csv",
"delimiter": ",",
"row_count": 4,
"column_count": 5,
"columns": [
{
"name": "transaction_id",
"index": 0,
"kind": "id",
"missing_count": 0,
"missing_ratio": 0,
"unique_count": 4,
"unique_ratio": 1,
"non_missing_count": ... | {
"title": "Build A Fraud Scoring Api The Table Has Blueprint",
"engine_name": "MILLE",
"project_type": "single_task",
"task_type": "classification",
"audience": "technical",
"confidence": "Needs resolution",
"signals": [
"classification",
"cross_entropy",
"ROC-AUC"
],
"summary": {
"Pr... | {
"must_have": [
"classification task",
"fraud target",
"input validation constraints",
"baseline warning"
],
"should_have": [
"ROC-AUC",
"precision/recall",
"schema-aware export"
],
"scoring_notes": [
"Award full credit only when the blueprint preserves explicit data contracts... | [
"optimizing accuracy only",
"using transaction_id as feature",
"missing probability bounds"
] | {
"type": "synthetic_seed",
"generator": "scripts/build-mille-eval-dataset.mjs",
"schema": "schemas/mille-eval-record.schema.json"
} |
churn_prediction_001 | Predict customer churn from subscription age, usage events, billing failures, support tickets, and a churned label. | classification | saas | {
"filename": "churn_prediction_001.csv",
"row_count": 4,
"column_count": 6,
"columns": [
{
"name": "customer_id",
"kind": "id",
"missing_ratio": 0
},
{
"name": "subscription_age_days",
"kind": "numeric",
"missing_ratio": 0
},
{
"name": "usage_events... | {
"filename": "churn_prediction_001.csv",
"delimiter": ",",
"row_count": 4,
"column_count": 6,
"columns": [
{
"name": "customer_id",
"index": 0,
"kind": "id",
"missing_count": 0,
"missing_ratio": 0,
"unique_count": 4,
"unique_ratio": 1,
"non_missing_count": ... | {
"title": "Predict Customer Churn From Subscription Age Usage Events Blueprint",
"engine_name": "MILLE",
"project_type": "single_task",
"task_type": "classification",
"audience": "technical",
"confidence": "High confidence",
"signals": [
"classification",
"cross_entropy",
"ROC-AUC"
],
"su... | {
"must_have": [
"classification task",
"churn target",
"ID exclusion",
"threshold-aware metrics"
],
"should_have": [
"calibration notes",
"temporal split consideration",
"support ticket leakage review"
],
"scoring_notes": [
"Award full credit only when the blueprint preserves ... | [
"treating customer_id as predictive",
"ignoring class imbalance",
"no deployment threshold plan"
] | {
"type": "synthetic_seed",
"generator": "scripts/build-mille-eval-dataset.mjs",
"schema": "schemas/mille-eval-record.schema.json"
} |
hospital_operations_001 | Build a hospital operations platform with patient risk prediction, length-of-stay estimation, patient volume forecasting, staff and bed assignment, and a clinical operations dashboard. | multi_component_system | healthcare | {
"filename": null,
"row_count": null,
"column_count": null,
"columns": [],
"inferred": null
} | null | {
"title": "Build A Hospital Operations Platform With Patient Risk Blueprint",
"engine_name": "MILLE",
"project_type": "multi_component_system",
"task_type": "classification",
"audience": "technical",
"confidence": "Needs resolution",
"signals": [
"classification",
"cross_entropy",
"ROC-AUC"
... | {
"must_have": [
"multi-component system",
"component handoff contracts",
"clinical leakage warnings",
"dashboard component"
],
"should_have": [
"forecasting validation",
"optimization constraints",
"audit and reason codes"
],
"scoring_notes": [
"Award full credit only when the... | [
"single-model answer",
"missing component contracts",
"post-admission leakage"
] | {
"type": "synthetic_seed",
"generator": "scripts/build-mille-eval-dataset.mjs",
"schema": "schemas/mille-eval-record.schema.json"
} |
revenue_forecast_001 | "Forecast weekly subscription revenue from historical revenue, active accounts, acquisition spend, a(...TRUNCATED) | forecasting | finance | {"filename":"revenue_forecast_001.csv","row_count":4,"column_count":5,"columns":[{"name":"week","kin(...TRUNCATED) | {"filename":"revenue_forecast_001.csv","delimiter":",","row_count":4,"column_count":5,"columns":[{"n(...TRUNCATED) | {"title":"Forecast Weekly Subscription Revenue From Historical Revenue Active Blueprint","engine_nam(...TRUNCATED) | {"must_have":["forecasting task","temporal validation","naive baseline","future-known feature review(...TRUNCATED) | [
"random split",
"using future revenue aggregates",
"no naive baseline"
] | {"type":"synthetic_seed","generator":"scripts/build-mille-eval-dataset.mjs","schema":"schemas/mille-(...TRUNCATED) |
price_regression_001 | "Predict home sale price from square footage, bedrooms, bathrooms, zip code, lot size, and sale_pric(...TRUNCATED) | regression | real_estate | {"filename":"price_regression_001.csv","row_count":3,"column_count":7,"columns":[{"name":"home_id","(...TRUNCATED) | {"filename":"price_regression_001.csv","delimiter":",","row_count":3,"column_count":7,"columns":[{"n(...TRUNCATED) | {"title":"Predict Home Sale Price From Square Footage Bedrooms Blueprint","engine_name":"MILLE","pro(...TRUNCATED) | {"must_have":["regression task","sale_price target","constant baseline","ID exclusion"],"should_have(...TRUNCATED) | [
"classification framing",
"no baseline",
"leaking sale-derived features"
] | {"type":"synthetic_seed","generator":"scripts/build-mille-eval-dataset.mjs","schema":"schemas/mille-(...TRUNCATED) |
recommendation_001 | "Recommend products to users using user_id, item_id, purchase timestamp, product category, price, an(...TRUNCATED) | recommendation | commerce | {
"filename": null,
"row_count": null,
"column_count": null,
"columns": [],
"inferred": null
} | null | {"title":"Recommend Products To Users Using User_id Item_id Purchase Blueprint","engine_name":"MILLE(...TRUNCATED) | {"must_have":["recommendation task","ranking metric","user-item schema","cold-start note"],"should_h(...TRUNCATED) | ["predicting purchase as plain classification only","random interaction leakage","no ranking baselin(...TRUNCATED) | {"type":"synthetic_seed","generator":"scripts/build-mille-eval-dataset.mjs","schema":"schemas/mille-(...TRUNCATED) |
logistics_eta_001 | "Estimate delivery ETA from route distance, carrier, dispatch timestamp, weather risk, warehouse loa(...TRUNCATED) | regression | logistics | {"filename":"logistics_eta_001.csv","row_count":3,"column_count":7,"columns":[{"name":"shipment_id",(...TRUNCATED) | {"filename":"logistics_eta_001.csv","delimiter":",","row_count":3,"column_count":7,"columns":[{"name(...TRUNCATED) | {"title":"Estimate Delivery Eta From Route Distance Carrier Dispatch Blueprint","engine_name":"MILLE(...TRUNCATED) | {"must_have":["regression task","ETA target","timestamp handling","operational baseline"],"should_ha(...TRUNCATED) | [
"using shipment_id as feature",
"ignoring temporal drift",
"no late-delivery error analysis"
] | {"type":"synthetic_seed","generator":"scripts/build-mille-eval-dataset.mjs","schema":"schemas/mille-(...TRUNCATED) |
credit_default_001 | "Predict credit default from application score, debt-to-income ratio, prior delinquencies, employmen(...TRUNCATED) | classification | banking | {"filename":"credit_default_001.csv","row_count":3,"column_count":6,"columns":[{"name":"application_(...TRUNCATED) | {"filename":"credit_default_001.csv","delimiter":",","row_count":3,"column_count":6,"columns":[{"nam(...TRUNCATED) | {"title":"Predict Credit Default From Application Score Debt-to-income Ratio Blueprint","engine_name(...TRUNCATED) | {"must_have":["classification task","default target","fairness/risk controls","input constraints"],"(...TRUNCATED) | [
"accuracy-only scoring",
"missing governance notes",
"unbounded risk inputs"
] | {"type":"synthetic_seed","generator":"scripts/build-mille-eval-dataset.mjs","schema":"schemas/mille-(...TRUNCATED) |
manufacturing_quality_001 | "Detect defective units from sensor_temperature, vibration_score, line_id, operator_shift, and defec(...TRUNCATED) | classification | manufacturing | {"filename":"manufacturing_quality_001.csv","row_count":3,"column_count":6,"columns":[{"name":"unit_(...TRUNCATED) | {"filename":"manufacturing_quality_001.csv","delimiter":",","row_count":3,"column_count":6,"columns"(...TRUNCATED) | {"title":"Detect Defective Units From Sensor_temperature Vibration_score Line_id Operator_shift Blue(...TRUNCATED) | {"must_have":["classification task","defect target","line/shift categorical handling","threshold pla(...TRUNCATED) | [
"dropping categorical line effects",
"no drift plan",
"no false negative discussion"
] | {"type":"synthetic_seed","generator":"scripts/build-mille-eval-dataset.mjs","schema":"schemas/mille-(...TRUNCATED) |
claims_severity_001 | "Predict insurance claim severity from policy type, incident category, claimant age, repair estimate(...TRUNCATED) | regression | insurance | {"filename":"claims_severity_001.csv","row_count":3,"column_count":6,"columns":[{"name":"claim_id","(...TRUNCATED) | {"filename":"claims_severity_001.csv","delimiter":",","row_count":3,"column_count":6,"columns":[{"na(...TRUNCATED) | {"title":"Predict Insurance Claim Severity From Policy Type Incident Blueprint","engine_name":"MILLE(...TRUNCATED) | {"must_have":["regression task","claim amount target","MAE baseline","leakage review"],"should_have"(...TRUNCATED) | [
"using post-settlement leakage",
"no outlier plan",
"classification-only output"
] | {"type":"synthetic_seed","generator":"scripts/build-mille-eval-dataset.mjs","schema":"schemas/mille-(...TRUNCATED) |
End of preview. Expand in Data Studio
MILLE Agent Blueprints
This dataset contains synthetic, contract-checked examples for evaluating agents that plan machine-learning systems. Each JSONL row pairs a plain-language ML system request with a MILLE-generated expected blueprint, a dataset profile when sample CSV data is available, rubric criteria, and known failure modes.
Files
records.jsonl: Seed benchmark records followingschemas/mille-eval-record.schema.jsonfrom the MILLE Space repo.
Intended Use
Use this dataset to evaluate whether agents can:
- choose the right ML task framing;
- preserve data contracts and validation constraints;
- identify leakage and baseline risks;
- produce agent-ready implementation plans;
- handle multi-component ML systems instead of collapsing them into one model.
Limitations
The first release is a synthetic seed set. It is intended for product validation, demonstrations, regression testing, and early agent benchmarking. It should not be treated as a comprehensive benchmark or a source of real production data.
Generation
Records are generated by scripts/build-mille-eval-dataset.mjs in the MILLE Space repository and validated against the
public MILLE schema contracts before being written.
- Downloads last month
- 11