harshbhx/ratchet-nemotron-nano-4b-text2sql
Text Generation • Updated • 66 • 2
prompt listlengths 2 2 | completion listlengths 1 1 | chat_template_kwargs dict | meta dict |
|---|---|---|---|
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT tail_number, model_type, seat_capacity FROM aero_horizon_fleet ORDER BY seat_capacity DESC LIMIT 5;"
}
] | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "simple",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT employee_number, first_name, last_name, role, hire_date FROM aero_horizon_crew WHERE date(REPLACE(hire_date, '/', '-')) > '2020-01-01' ORDER BY hire_date;"
}
] | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "simple",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT route_code, origin_hub_id, destination_hub_id, distance_km FROM aero_horizon_routes WHERE is_seasonal = 1;"
}
] | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "simple",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT f.flight_number, r.route_code, oh.city AS origin_city, dh.city AS destination_city FROM aero_horizon_flights f JOIN aero_horizon_routes r ON f.route_id = r.route_id JOIN aero_horizon_hubs oh ON r.origin_hub_id = oh.hub_id JOIN aero_horizon_hubs dh ON r.destination_h... | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "join",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT f.tail_number, f.model_type, h.hub_name, h.city FROM aero_horizon_fleet f JOIN aero_horizon_hubs h ON f.hub_id = h.hub_id WHERE f.status = 'Active';"
}
] | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "join",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT f.flight_number, c.first_name || ' ' || c.last_name AS captain_name, ac.tail_number FROM aero_horizon_flights f JOIN aero_horizon_crew c ON f.captain_id = c.crew_id JOIN aero_horizon_fleet ac ON f.aircraft_id = ac.aircraft_id WHERE f.flight_status = 'Arrived';"
}
... | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "join",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT model_type, COUNT(*) AS active_count FROM aero_horizon_fleet WHERE status = 'Active' GROUP BY model_type;"
}
] | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "aggregate",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT r.route_code, AVG(f.passenger_count) AS avg_passengers, AVG(f.cargo_weight_kg) AS avg_cargo_kg FROM aero_horizon_flights f JOIN aero_horizon_routes r ON f.route_id = r.route_id WHERE f.flight_status = 'Arrived' GROUP BY r.route_code;"
}
] | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "aggregate",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT r.route_code, r.scheduled_duration_min, AVG(f.actual_duration_min) AS avg_actual_duration FROM aero_horizon_flights f JOIN aero_horizon_routes r ON f.route_id = r.route_id WHERE f.flight_status = 'Arrived' AND f.actual_duration_min IS NOT NULL GROUP BY r.route_code,... | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "hard",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n... | [
{
"role": "assistant",
"content": "SELECT c.first_name || ' ' || c.last_name AS captain_name, COUNT(DISTINCT f.model_type) AS model_types_flown FROM aero_horizon_flights fl JOIN aero_horizon_crew c ON fl.captain_id = c.crew_id JOIN aero_horizon_fleet f ON fl.aircraft_id = f.aircraft_id WHERE fl.flight_statu... | {
"enable_thinking": false
} | {
"db": "test-0",
"domain": "airline operations",
"difficulty": "hard",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY,\n company_name TEXT NOT NULL,\n co... | [
{
"role": "assistant",
"content": "SELECT product_name, selling_price FROM products ORDER BY selling_price DESC LIMIT 5;"
}
] | {
"enable_thinking": false
} | {
"db": "test-1",
"domain": "pharmacy inventory",
"difficulty": "simple",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY,\n company_name TEXT NOT NULL,\n co... | [
{
"role": "assistant",
"content": "SELECT p.product_name, p.sku FROM products p JOIN suppliers s ON p.supplier_id = s.supplier_id WHERE s.company_name = 'Apex Pharmaceuticals';"
}
] | {
"enable_thinking": false
} | {
"db": "test-1",
"domain": "pharmacy inventory",
"difficulty": "simple",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY,\n company_name TEXT NOT NULL,\n co... | [
{
"role": "assistant",
"content": "SELECT p.product_name, ib.batch_number, ib.quantity_on_hand, ib.expiry_date FROM inventory_batches ib JOIN products p ON ib.product_id = p.product_id ORDER BY p.product_name, ib.expiry_date;"
}
] | {
"enable_thinking": false
} | {
"db": "test-1",
"domain": "pharmacy inventory",
"difficulty": "join",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY,\n company_name TEXT NOT NULL,\n co... | [
{
"role": "assistant",
"content": "SELECT st.transaction_type, st.quantity_change, st.transaction_date FROM stock_transactions st JOIN inventory_batches ib ON st.batch_id = ib.batch_id JOIN products p ON ib.product_id = p.product_id WHERE p.product_name = 'Amoxicillin 500mg Capsule' ORDER BY st.transaction_... | {
"enable_thinking": false
} | {
"db": "test-1",
"domain": "pharmacy inventory",
"difficulty": "join",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY,\n company_name TEXT NOT NULL,\n co... | [
{
"role": "assistant",
"content": "SELECT p.product_name, s.company_name FROM products p JOIN suppliers s ON p.supplier_id = s.supplier_id WHERE s.is_active = 0;"
}
] | {
"enable_thinking": false
} | {
"db": "test-1",
"domain": "pharmacy inventory",
"difficulty": "join",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY,\n company_name TEXT NOT NULL,\n co... | [
{
"role": "assistant",
"content": "SELECT pc.category_name, SUM(ib.quantity_on_hand) AS total_quantity FROM inventory_batches ib JOIN products p ON ib.product_id = p.product_id JOIN product_categories pc ON p.category_id = pc.category_id GROUP BY pc.category_name ORDER BY total_quantity DESC;"
}
] | {
"enable_thinking": false
} | {
"db": "test-1",
"domain": "pharmacy inventory",
"difficulty": "aggregate",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY,\n company_name TEXT NOT NULL,\n co... | [
{
"role": "assistant",
"content": "SELECT p.product_name, ABS(SUM(st.quantity_change)) AS units_sold FROM stock_transactions st JOIN inventory_batches ib ON st.batch_id = ib.batch_id JOIN products p ON ib.product_id = p.product_id WHERE st.transaction_type = 'SALE' AND strftime('%Y', REPLACE(st.transaction_... | {
"enable_thinking": false
} | {
"db": "test-1",
"domain": "pharmacy inventory",
"difficulty": "aggregate",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY,\n company_name TEXT NOT NULL,\n co... | [
{
"role": "assistant",
"content": "SELECT p.product_name, p.reorder_level, COALESCE(SUM(ib.quantity_on_hand), 0) AS total_stock FROM products p LEFT JOIN inventory_batches ib ON p.product_id = ib.product_id GROUP BY p.product_id, p.product_name, p.reorder_level HAVING total_stock < p.reorder_level;"
}
] | {
"enable_thinking": false
} | {
"db": "test-1",
"domain": "pharmacy inventory",
"difficulty": "hard",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY,\n company_name TEXT NOT NULL,\n co... | [
{
"role": "assistant",
"content": "SELECT s.company_name, SUM(ib.quantity_on_hand * ib.unit_cost_at_purchase) AS total_inventory_value FROM inventory_batches ib JOIN products p ON ib.product_id = p.product_id JOIN suppliers s ON p.supplier_id = s.supplier_id WHERE date(REPLACE(ib.expiry_date, '/', '-')) >= ... | {
"enable_thinking": false
} | {
"db": "test-1",
"domain": "pharmacy inventory",
"difficulty": "hard",
"domain_seen": false
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE branches (\n branch_id INTEGER PRIMARY KEY,\n branch_name TEXT NOT NULL,\n city T... | [
{
"role": "assistant",
"content": "SELECT room_id, room_number, floor_level, has_balcony FROM accommodations WHERE status = 'Occupied' ORDER BY room_id;"
}
] | {
"enable_thinking": false
} | {
"db": "test-3",
"domain": "hotel reservations",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE branches (\n branch_id INTEGER PRIMARY KEY,\n branch_name TEXT NOT NULL,\n city T... | [
{
"role": "assistant",
"content": "SELECT guest_id, full_name, email, loyalty_tier, registration_date FROM guest_profiles ORDER BY registration_date DESC LIMIT 5;"
}
] | {
"enable_thinking": false
} | {
"db": "test-3",
"domain": "hotel reservations",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE branches (\n branch_id INTEGER PRIMARY KEY,\n branch_name TEXT NOT NULL,\n city T... | [
{
"role": "assistant",
"content": "SELECT category_id, category_name, base_rate, max_occupancy FROM room_categories WHERE base_rate > 300 ORDER BY base_rate;"
}
] | {
"enable_thinking": false
} | {
"db": "test-3",
"domain": "hotel reservations",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE branches (\n branch_id INTEGER PRIMARY KEY,\n branch_name TEXT NOT NULL,\n city T... | [
{
"role": "assistant",
"content": "SELECT gp.full_name, rc.category_name, b.city\nFROM bookings bk\nJOIN guest_profiles gp ON bk.guest_id = gp.guest_id\nJOIN accommodations a ON bk.room_id = a.room_id\nJOIN room_categories rc ON a.category_id = rc.category_id\nJOIN branches b ON a.branch_id = b.branch_id\nW... | {
"enable_thinking": false
} | {
"db": "test-3",
"domain": "hotel reservations",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE branches (\n branch_id INTEGER PRIMARY KEY,\n branch_name TEXT NOT NULL,\n city T... | [
{
"role": "assistant",
"content": "SELECT a.room_id, a.room_number, rc.category_name, br.branch_name, br.city\nFROM accommodations a\nJOIN room_categories rc ON a.category_id = rc.category_id\nJOIN branches br ON a.branch_id = br.branch_id\nWHERE a.status = 'Available' AND a.has_balcony = 1\nORDER BY br.bra... | {
"enable_thinking": false
} | {
"db": "test-3",
"domain": "hotel reservations",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE branches (\n branch_id INTEGER PRIMARY KEY,\n branch_name TEXT NOT NULL,\n city T... | [
{
"role": "assistant",
"content": "SELECT gp.full_name, gp.loyalty_tier, rc.category_name, bk.check_in_date\nFROM bookings bk\nJOIN guest_profiles gp ON bk.guest_id = gp.guest_id\nJOIN accommodations a ON bk.room_id = a.room_id\nJOIN room_categories rc ON a.category_id = rc.category_id\nWHERE bk.booking_sta... | {
"enable_thinking": false
} | {
"db": "test-3",
"domain": "hotel reservations",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE branches (\n branch_id INTEGER PRIMARY KEY,\n branch_name TEXT NOT NULL,\n city T... | [
{
"role": "assistant",
"content": "SELECT br.branch_name, rc.category_name, COUNT(a.room_id) AS room_count\nFROM accommodations a\nJOIN branches br ON a.branch_id = br.branch_id\nJOIN room_categories rc ON a.category_id = rc.category_id\nGROUP BY br.branch_id, rc.category_id\nORDER BY br.branch_name, rc.cat... | {
"enable_thinking": false
} | {
"db": "test-3",
"domain": "hotel reservations",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE branches (\n branch_id INTEGER PRIMARY KEY,\n branch_name TEXT NOT NULL,\n city T... | [
{
"role": "assistant",
"content": "SELECT br.branch_name,\n COUNT(bk.booking_id) AS completed_stays,\n SUM(bk.total_charge) AS total_revenue,\n AVG(bk.total_charge) AS avg_charge\nFROM bookings bk\nJOIN accommodations a ON bk.room_id = a.room_id\nJOIN branches br ON a.branch_id = br.branch... | {
"enable_thinking": false
} | {
"db": "test-3",
"domain": "hotel reservations",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE branches (\n branch_id INTEGER PRIMARY KEY,\n branch_name TEXT NOT NULL,\n city T... | [
{
"role": "assistant",
"content": "SELECT rc.category_name,\n COUNT(a.room_id) AS total_rooms,\n SUM(CASE WHEN a.status = 'Occupied' THEN 1 ELSE 0 END) AS occupied_rooms,\n ROUND(100.0 * SUM(CASE WHEN a.status = 'Occupied' THEN 1 ELSE 0 END) / COUNT(a.room_id), 2) AS occupancy_rate_pct\nFR... | {
"enable_thinking": false
} | {
"db": "test-3",
"domain": "hotel reservations",
"difficulty": "hard",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE medical_facilities (\n facility_id INTEGER PRIMARY KEY,\n facility_name TEXT NOT NUL... | [
{
"role": "assistant",
"content": "SELECT facility_name, city, state FROM medical_facilities WHERE is_active = 1 ORDER BY facility_name;"
}
] | {
"enable_thinking": false
} | {
"db": "test-4",
"domain": "clinic appointments",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE medical_facilities (\n facility_id INTEGER PRIMARY KEY,\n facility_name TEXT NOT NUL... | [
{
"role": "assistant",
"content": "SELECT first_name, last_name, medical_specialty, hire_date FROM care_providers WHERE employment_status = 'Active' ORDER BY hire_date DESC LIMIT 5;"
}
] | {
"enable_thinking": false
} | {
"db": "test-4",
"domain": "clinic appointments",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE medical_facilities (\n facility_id INTEGER PRIMARY KEY,\n facility_name TEXT NOT NUL... | [
{
"role": "assistant",
"content": "SELECT visit_id, appointment_datetime, visit_reason, visit_status FROM scheduled_visits WHERE strftime('%Y-%m', REPLACE(appointment_datetime, '/', '-')) = '2024-02' ORDER BY appointment_datetime;"
}
] | {
"enable_thinking": false
} | {
"db": "test-4",
"domain": "clinic appointments",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE medical_facilities (\n facility_id INTEGER PRIMARY KEY,\n facility_name TEXT NOT NUL... | [
{
"role": "assistant",
"content": "SELECT rp.first_name || ' ' || rp.last_name AS patient_name, cp.first_name || ' ' || cp.last_name AS provider_name, cp.medical_specialty FROM registered_patients rp JOIN care_providers cp ON rp.facility_id = cp.facility_id WHERE cp.employment_status = 'Active' ORDER BY pat... | {
"enable_thinking": false
} | {
"db": "test-4",
"domain": "clinic appointments",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE medical_facilities (\n facility_id INTEGER PRIMARY KEY,\n facility_name TEXT NOT NUL... | [
{
"role": "assistant",
"content": "SELECT rp.first_name || ' ' || rp.last_name AS patient_name, cp.first_name || ' ' || cp.last_name AS provider_name, mf.facility_name, sv.appointment_datetime, sv.visit_reason FROM scheduled_visits sv JOIN registered_patients rp ON sv.patient_id = rp.patient_id JOIN care_pr... | {
"enable_thinking": false
} | {
"db": "test-4",
"domain": "clinic appointments",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE medical_facilities (\n facility_id INTEGER PRIMARY KEY,\n facility_name TEXT NOT NUL... | [
{
"role": "assistant",
"content": "SELECT mf.facility_name, cp.first_name || ' ' || cp.last_name AS provider_name, cp.medical_specialty FROM medical_facilities mf JOIN care_providers cp ON mf.facility_id = cp.facility_id WHERE mf.is_active = 1 AND cp.employment_status = 'Active' ORDER BY mf.facility_name, p... | {
"enable_thinking": false
} | {
"db": "test-4",
"domain": "clinic appointments",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE medical_facilities (\n facility_id INTEGER PRIMARY KEY,\n facility_name TEXT NOT NUL... | [
{
"role": "assistant",
"content": "SELECT visit_type, COUNT(*) AS visit_count FROM scheduled_visits WHERE visit_status = 'Completed' GROUP BY visit_type ORDER BY visit_count DESC;"
}
] | {
"enable_thinking": false
} | {
"db": "test-4",
"domain": "clinic appointments",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE medical_facilities (\n facility_id INTEGER PRIMARY KEY,\n facility_name TEXT NOT NUL... | [
{
"role": "assistant",
"content": "SELECT mf.facility_name, ROUND(AVG(sv.copay_amount), 2) AS avg_copay FROM scheduled_visits sv JOIN medical_facilities mf ON sv.facility_id = mf.facility_id WHERE sv.visit_status = 'Completed' GROUP BY mf.facility_name ORDER BY avg_copay DESC;"
}
] | {
"enable_thinking": false
} | {
"db": "test-4",
"domain": "clinic appointments",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE medical_facilities (\n facility_id INTEGER PRIMARY KEY,\n facility_name TEXT NOT NUL... | [
{
"role": "assistant",
"content": "SELECT cp.first_name || ' ' || cp.last_name AS provider_name, cp.medical_specialty, COUNT(sv.visit_id) AS completed_visits, SUM(sv.copay_amount) AS total_copay FROM care_providers cp JOIN scheduled_visits sv ON cp.provider_id = sv.provider_id WHERE sv.visit_status = 'Compl... | {
"enable_thinking": false
} | {
"db": "test-4",
"domain": "clinic appointments",
"difficulty": "hard",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT agent_id, full_name, hire_date FROM support_agents WHERE is_active = 1 ORDER BY date(REPLACE(hire_date, '/', '-')) DESC LIMIT 5;"
}
] | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT ticket_id, subject, priority, status FROM support_tickets WHERE status IN ('Open', 'In Progress') ORDER BY CASE priority WHEN 'Critical' THEN 1 WHEN 'High' THEN 2 WHEN 'Medium' THEN 3 WHEN 'Low' THEN 4 END;"
}
] | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT product_id, product_name, category FROM product_lines WHERE is_legacy = 1;"
}
] | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT st.ticket_id, st.subject, ca.company_name, pl.product_name FROM support_tickets st JOIN customer_accounts ca ON st.account_id = ca.account_id JOIN product_lines pl ON st.product_id = pl.product_id;"
}
] | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT ta.activity_id, sa.full_name, ta.activity_type, ta.description, ta.duration_minutes FROM ticket_activities ta JOIN support_agents sa ON ta.agent_id = sa.agent_id WHERE ta.ticket_id = 1 ORDER BY ta.created_at;"
}
] | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT sa.full_name, COUNT(st.ticket_id) AS assigned_tickets FROM support_agents sa LEFT JOIN support_tickets st ON sa.agent_id = st.assigned_agent_id AND st.status NOT IN ('Resolved', 'Closed') GROUP BY sa.agent_id, sa.full_name ORDER BY assigned_tickets DESC;"
}
] | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT subscription_tier, ROUND(AVG(monthly_mrr), 2) AS avg_mrr FROM customer_accounts GROUP BY subscription_tier ORDER BY avg_mrr DESC;"
}
] | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT pl.product_name, COUNT(st.ticket_id) AS total_tickets, ROUND(AVG(st.satisfaction_score), 2) AS avg_satisfaction FROM support_tickets st JOIN product_lines pl ON st.product_id = pl.product_id WHERE st.satisfaction_score IS NOT NULL GROUP BY pl.product_id, pl.product_... | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT sa.full_name, SUM(ta.duration_minutes) AS total_customer_minutes FROM ticket_activities ta JOIN support_agents sa ON ta.agent_id = sa.agent_id WHERE ta.is_internal = 0 GROUP BY sa.agent_id, sa.full_name HAVING total_customer_minutes > 100 ORDER BY total_customer_min... | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "hard",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE support_agents (\n agent_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ema... | [
{
"role": "assistant",
"content": "SELECT ca.company_name, ca.subscription_tier, COUNT(st.ticket_id) AS critical_high_tickets, CASE WHEN ca.subscription_tier = 'Enterprise' THEN 'Yes' ELSE 'No' END AS is_enterprise FROM customer_accounts ca LEFT JOIN support_tickets st ON ca.account_id = st.account_id AND s... | {
"enable_thinking": false
} | {
"db": "test-7",
"domain": "customer support desk",
"difficulty": "hard",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT track_name, standard_fee FROM certification_track ORDER BY standard_fee DESC;"
}
] | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT full_name, specialization, hourly_rate FROM facilitator WHERE is_active = 0;"
}
] | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT full_name, hire_date FROM facilitator ORDER BY hire_date DESC LIMIT 5;"
}
] | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT ct.track_name, c.cohort_label, f.full_name AS lead_facilitator, COUNT(r.registration_id) AS enrolled_count\nFROM cohort c\nJOIN certification_track ct ON c.track_id = ct.track_id\nJOIN facilitator f ON c.lead_facilitator_id = f.facilitator_id\nLEFT JOIN registration... | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT l.full_name, l.department, c.cohort_label\nFROM learner l\nJOIN registration r ON l.learner_id = r.learner_id\nJOIN cohort c ON r.cohort_id = c.cohort_id\nJOIN certification_track ct ON c.track_id = ct.track_id\nWHERE ct.track_code = 'CDP-200';"
}
] | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT f.full_name AS facilitator, ss.session_date, ss.topic, CASE ss.is_mandatory WHEN 1 THEN 'Yes' ELSE 'No' END AS mandatory\nFROM session_schedule ss\nJOIN facilitator f ON ss.facilitator_id = f.facilitator_id\nJOIN cohort c ON ss.cohort_id = c.cohort_id\nWHERE c.cohor... | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT c.cohort_label, ROUND(AVG(r.final_score), 2) AS avg_score, ROUND(AVG(r.attendance_pct), 2) AS avg_attendance\nFROM registration r\nJOIN cohort c ON r.cohort_id = c.cohort_id\nWHERE c.status = 'Completed' AND r.final_score IS NOT NULL\nGROUP BY c.cohort_id, c.cohort_... | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT f.full_name, f.specialization, COUNT(DISTINCT ss.cohort_id) AS cohort_count\nFROM facilitator f\nJOIN session_schedule ss ON f.facilitator_id = ss.facilitator_id\nGROUP BY f.facilitator_id, f.full_name, f.specialization\nHAVING cohort_count > 0\nORDER BY cohort_coun... | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT l.full_name, \n ct_completed.track_name AS completed_program,\n ct_active.track_name AS active_program\nFROM learner l\nJOIN registration r_comp ON l.learner_id = r_comp.learner_id\nJOIN cohort c_comp ON r_comp.cohort_id = c_comp.cohort_id\nJOIN certific... | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "hard",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE facilitator (\n facilitator_id INTEGER PRIMARY KEY,\n full_name TEXT NOT NULL,\n ... | [
{
"role": "assistant",
"content": "SELECT ct.track_name,\n SUM(ct.standard_fee) AS total_revenue,\n ROUND(AVG(r.final_score), 2) AS avg_paid_score\nFROM certification_track ct\nJOIN cohort c ON ct.track_id = c.track_id\nJOIN registration r ON c.cohort_id = r.cohort_id\nWHERE c.status = 'Complete... | {
"enable_thinking": false
} | {
"db": "test-8",
"domain": "university courses",
"difficulty": "hard",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE game_project (\n project_id INTEGER PRIMARY KEY,\n title TEXT NOT NULL,\n genre T... | [
{
"role": "assistant",
"content": "SELECT project_id, title, budget_usd FROM game_project WHERE budget_usd > 2000000 ORDER BY budget_usd DESC;"
}
] | {
"enable_thinking": false
} | {
"db": "test-9",
"domain": "video game studio",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE game_project (\n project_id INTEGER PRIMARY KEY,\n title TEXT NOT NULL,\n genre T... | [
{
"role": "assistant",
"content": "SELECT member_id, full_name, role_title, hire_date FROM studio_member WHERE is_active = 1 AND date(REPLACE(hire_date, '/', '-')) > '2022-01-01' ORDER BY hire_date;"
}
] | {
"enable_thinking": false
} | {
"db": "test-9",
"domain": "video game studio",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE game_project (\n project_id INTEGER PRIMARY KEY,\n title TEXT NOT NULL,\n genre T... | [
{
"role": "assistant",
"content": "SELECT asset_name, asset_type, project_id, file_size_mb FROM asset ORDER BY file_size_mb DESC LIMIT 5;"
}
] | {
"enable_thinking": false
} | {
"db": "test-9",
"domain": "video game studio",
"difficulty": "simple",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE game_project (\n project_id INTEGER PRIMARY KEY,\n title TEXT NOT NULL,\n genre T... | [
{
"role": "assistant",
"content": "SELECT gp.title, sm.full_name FROM game_project gp JOIN studio_member sm ON gp.lead_producer_id = sm.member_id;"
}
] | {
"enable_thinking": false
} | {
"db": "test-9",
"domain": "video game studio",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE game_project (\n project_id INTEGER PRIMARY KEY,\n title TEXT NOT NULL,\n genre T... | [
{
"role": "assistant",
"content": "SELECT a.asset_name, a.asset_type, sm.full_name AS artist_name, gp.title AS project_title FROM asset a JOIN studio_member sm ON a.assigned_artist_id = sm.member_id JOIN game_project gp ON a.project_id = gp.project_id WHERE a.review_status = 'In Review';"
}
] | {
"enable_thinking": false
} | {
"db": "test-9",
"domain": "video game studio",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE game_project (\n project_id INTEGER PRIMARY KEY,\n title TEXT NOT NULL,\n genre T... | [
{
"role": "assistant",
"content": "SELECT m.name AS milestone_name, gp.title AS project_title, m.planned_date, m.status FROM milestone m JOIN game_project gp ON m.project_id = gp.project_id WHERE m.status IN ('Delayed', 'In Progress') ORDER BY m.planned_date;"
}
] | {
"enable_thinking": false
} | {
"db": "test-9",
"domain": "video game studio",
"difficulty": "join",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE game_project (\n project_id INTEGER PRIMARY KEY,\n title TEXT NOT NULL,\n genre T... | [
{
"role": "assistant",
"content": "SELECT gp.title, COUNT(a.asset_id) AS asset_count, ROUND(AVG(a.file_size_mb), 2) AS avg_file_size_mb FROM game_project gp LEFT JOIN asset a ON gp.project_id = a.project_id GROUP BY gp.project_id, gp.title ORDER BY asset_count DESC;"
}
] | {
"enable_thinking": false
} | {
"db": "test-9",
"domain": "video game studio",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE game_project (\n project_id INTEGER PRIMARY KEY,\n title TEXT NOT NULL,\n genre T... | [
{
"role": "assistant",
"content": "SELECT gp.title, COUNT(sm.member_id) AS active_member_count, ROUND(AVG(sm.hourly_rate_usd), 2) AS avg_hourly_rate FROM game_project gp LEFT JOIN studio_member sm ON gp.project_id = sm.project_id AND sm.is_active = 1 GROUP BY gp.project_id, gp.title ORDER BY active_member_c... | {
"enable_thinking": false
} | {
"db": "test-9",
"domain": "video game studio",
"difficulty": "aggregate",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE game_project (\n project_id INTEGER PRIMARY KEY,\n title TEXT NOT NULL,\n genre T... | [
{
"role": "assistant",
"content": "SELECT ec.company_name, COUNT(DISTINCT ec.project_id) AS project_count, GROUP_CONCAT(gp.title, ', ') AS project_titles FROM external_contractor ec JOIN game_project gp ON ec.project_id = gp.project_id GROUP BY ec.company_name HAVING project_count > 1;"
}
] | {
"enable_thinking": false
} | {
"db": "test-9",
"domain": "video game studio",
"difficulty": "hard",
"domain_seen": true
} |
[
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE suppliers (\n supplier_id INTEGER PRIMARY KEY AUTOINCREMENT,\n supplier_name TEXT NO... | [
{
"role": "assistant",
"content": "SELECT generic_name, brand_name, strength, dosage_form, unit_cost FROM medications WHERE is_active = 1 AND unit_cost > 0.50 ORDER BY unit_cost DESC;"
}
] | {
"enable_thinking": false
} | {
"db": "test-11",
"domain": "pharmacy inventory",
"difficulty": "simple",
"domain_seen": false
} |
End of preview. Expand in Data Studio
Ratchet text-to-SQL exams: questions, answers and scores
Two held-out exams for text-to-SQL, with every model's answers and every score, so each number reported for the Ratchet adapter (harshbhx/ratchet-nemotron-nano-4b-text2sql) can be recomputed on a laptop with no GPU. Part of Ratchet, built for the Nebius x NVIDIA Global AI Hackathon.
Example row
First row of exam1/golden.jsonl, without the answer key (completion). Long strings are shortened with ....
{
"prompt": [
{
"role": "system",
"content": "You are a SQLite expert. Given a database schema and a question, reply with one SQLite SELECT query and nothing else."
},
{
"role": "user",
"content": "Schema:\nCREATE TABLE aero_horizon_hubs (\n hub_id INTEGER PRIMARY KEY,\n hub_code TEXT NOT NULL UNIQUE,\n hub_name TEXT NOT NULL,\n city TEXT NOT NULL,\n country TEXT NOT NULL,\n timezone_offset INTEGER NOT NULL,\n is_international INTEGER NOT NULL CHECK (is_international IN (0, 1))\n);\n\nCREATE TABLE aero_horizon_fleet (\n aircraft_id INTEGER PRIMARY KEY,\n tail_number TEXT NOT NULL UNIQUE,\n model ..."
}
],
"chat_template_kwargs": {
"enable_thinking": false
},
"meta": {
"db": "test-0",
"domain": "airline operations",
"difficulty": "simple",
"domain_seen": false
}
}
These are exams: a test set to evaluate on, never to train on. Please do not train on them (exam 2 was written by Anthropic's Claude Sonnet 5.5, and the whole dataset carries an evaluation-only notice, see License): it would make any score on them meaningless. The training data is separate (harshbhx/ratchet-text2sql-synthetic) and was checked not to overlap with either exam.
The two exams and who wrote them
| Exam 1 | Exam 2 | |
|---|---|---|
| Questions | 172 | 200 |
| Written | 22 Oct 2026 | 23 Oct 2026 |
| Written by | NVIDIA Nemotron 3 Ultra, the same model that wrote the training data | Anthropic Claude Sonnet 5.5, a different lab's model, for evaluation only |
| Industries | 14; 57 questions are from five industries never used in training | 10 industries in neither the training data nor exam 1: museum collections, veterinary clinic, wine distribution, film production, wind farm maintenance, court cases, recycling plant, podcast network, cinema chain, fishing fleet |
| Mix | 51 simple, 53 join, 36 aggregate, 32 hard | 60 simple, 60 join, 40 aggregate, 40 hard |
Fingerprint (SHA-256 of golden.jsonl) |
6b9c1db49bbe2aecf4157ed27845fe024d1b4fcb540f93d81b6987b982b22864 |
b7634971cf6b0d79e64e30e4c16d1253fad65ab9a501a06edc6c9f8e70d69058 |
Exam 1 was written by the teacher, which flatters the teacher and anything trained to copy it. That is why exam 2 exists: a different author, written after training, on new industries. Exam 2's fingerprint was recorded when it was generated, before any model answered it (exam2/exam_manifest.json).
How each exam was checked. Every reference query was run on its database in a sandbox and kept only if it was one read-only SELECT returning 1 to 100 rows. No question, query or set of table names is shared with the training data.
How these exams were made
- Exam 1 (172 questions). NVIDIA Nemotron 3 Ultra (through Nebius Token Factory) wrote the databases, the questions and the reference queries with
scripts/gen_sql_data.py, on 22 Oct 2026. A script then ran every reference query in a sandbox and kept only those that were one read-onlySELECTreturning 1 to 100 rows. The project records say Claude was not used to write or check this exam. No person reviewed the rows. - Exam 2 (200 questions). Anthropic's Claude Sonnet 5.5, through the Anthropic API, wrote the 10 databases, the questions and the reference queries with
scripts/gen_exam2.py, on 23 Oct 2026 (10 calls; its raw responses are inexam2/writer_raw_responses/). The same script ran every reference query and kept only the ones that passed the checks above (it rejected none). No person reviewed the rows. - Answers. The files in
answers/are outputs of the models named in their file names (the base model, the two adapters, and Nemotron 3 Ultra as the teacher).results/holds scorer output, computed by a script. - Not fully recorded. The raw API responses of exam 1's original generation were not kept (only the saved text and token totals), so provenance for exam 1 rests on the script and the project records, not on a response log. Because of that, and because the two exams sit in one dataset, the evaluation-only notice below is applied to the whole dataset, not only to exam 2.
Scoring (strict)
An answer is right when running it on the question's database returns exactly the same rows as the reference query (column names ignored; row order only if the reference query orders). One reference answer per question, so a correct query that returns an extra column counts as wrong. This is deliberately strict and it penalises models that write wide answers: the teacher's low score below is mostly of that kind.
Right answers out of the exam, with the 95% range (Wilson):
| Exam 1 (172 questions, written by the teacher) | Exam 2 (200 questions, written by a different model) | |
|---|---|---|
| Base model, no adapter | 22 / 172 = 12.8% (8.6 to 18.6) | 25 / 200 = 12.5% (8.6 to 17.8) |
| Teacher (Nemotron 3 Ultra), same prompts | 44 / 172 = 25.6% (19.6 to 32.6) | 30 / 200 = 15.0% (10.7 to 20.6) |
| Adapter round 1 (821 training rows) | 57 / 172 = 33.1% (26.5 to 40.5) | 56 / 200 = 28.0% (22.2 to 34.6) |
| Adapter round 2 (4,322 training rows) | 67 / 172 = 39.0% (32.0 to 46.4) | 66 / 200 = 33.0% (26.9 to 39.8) |
- Round 2 beat round 1 by 10 questions on both exams (p = 0.121 and 0.132): consistent in direction, not proven by either exam alone.
- The adapter does not replace the teacher: at a 95% precision target its confidence gate passes almost none of the questions. The exam's one-reference rule favours short answers.
- All answering used one greedy answer per question, thinking off, at most 200 new tokens (the teacher: up to 600). The two adapters and the base model answered on a borrowed GPU; the teacher on Nebius Token Factory (about $0.14 and $0.11 for the two exams).
Files
| Path | What |
|---|---|
exam1/golden.jsonl, exam2/golden.jsonl |
The questions (a prompt with the schema and question; completion is the reference query; meta has the database id, industry and difficulty) |
exam1/dbs.json, exam2/dbs.json |
The SQL scripts that build each database |
exam2/exam_manifest.json, exam2/writer_raw_responses/ |
How exam 2 was generated: the author model, token counts, rejects, and the author's full raw responses |
answers/ |
Every model's raw answers on both exams, including round 1's answers under a second prompt wording |
results/ |
The scorer's output for each answer file, its verify outputs, the teacher checks, the leakage check and exams_summary.json (every number on the Ratchet pages) |
scripts/ |
The scorer (measure_adapter.py), the exam writer (gen_exam2.py), the summary script (exam_numbers.py), the teacher check and the training-data generator |
LICENSE.OpenMDW-1.1.txt |
The license of the model that wrote exam 1 and the training data, for reference |
MANIFEST.json |
SHA-256 of every file |
Check it yourself
python scripts/measure_adapter.py verify \
--raw answers/exam2_round2_base_and_adapter.jsonl \
--results results/exam2_round2.json \
--golden exam2/golden.jsonl --dbs exam2/dbs.json
# OK: every number in results.json was recomputed from raw.jsonl and matches
The verify command recomputes every number in a results file from the raw answers; it needs no GPU and no network.
Limitations
- Synthetic databases and questions written by language models; real databases are messier.
- One reference query per question; small exams (a score is good to about 7 points either way).
- English and SQLite only.
- Exam 2 is written by one model; another author would give somewhat different questions.
Licenses and provenance
- Evaluation only.
license: other: seeLICENSE-EVAL-ONLY.md. Read the rows and use them to evaluate models; do not use them to train, fine-tune or distill any model. This project never did. Exam 2 was written with Claude and the Anthropic terms restrict training on its outputs; to stay on the safe side the notice covers the whole dataset. - The code is Apache-2.0. The scripts in
scripts/(the scorer, the exam writer, the summary script, the teacher check and the data generator) are released under the Apache License 2.0. - Exam 1 and the databases of the training data were generated by NVIDIA Nemotron 3 Ultra, distributed under the OpenMDW License Agreement 1.1, which "does not impose any restrictions or obligations with respect to any use, modification, or sharing of any outputs generated by using the Model Materials". A copy of that license is included for reference. Taken alone, exam 1 would not need the evaluation-only restriction; it is applied because of the points under "How these exams were made".
- Exam 2 was generated by Anthropic's Claude Sonnet 5.5 through the Anthropic API. Under Anthropic's Commercial Terms (effective 17 June 2025) the customer owns the outputs, subject to the terms, and the terms bar using the Services to build competing products or to train competing AI models. Here the outputs are used only to evaluate models, never to train any; we ask that you do the same. Every reference query was executed before sharing, but they may still contain errors, and no person checked them.
- No exam content was written by a person, or by the base model or the adapters that are scored on it. Exam 1 was written by the teacher, which is also one of the scored models (its score is reported, and the exam favours it); exam 2 was written by a model that is not scored.
- Answers in
answers/were produced by the models named in their file names, including the base model NVIDIA Nemotron 3 Nano 4B (NVIDIA Nemotron Open Model License) with the adapters inharshbhx/ratchet-nemotron-nano-4b-text2sql. Their texts are short SQL queries. - Not affiliated with or endorsed by NVIDIA, Nebius or Anthropic.
Citation
@misc{ratchet2026,
author = {Harsh Bhanushali},
title = {Ratchet: license-clean distillation on open models},
year = {2026},
url = {https://github.com/harshbhx/ratchet}
}
- Downloads last month
- 209
Models trained or fine-tuned on harshbhx/ratchet-text2sql-exams