# bird-bench / debit_card_specializing__1481 - taskset: [bird-bench](https://harnessreport.com/tasks/bird-bench.md) - difficulty: hard - category: database - language: - runnable from the site: no - agent timeout: 3600s ## Results by harness _none yet_ ## Instruction ``` You are given a natural language question and a SQLite database. Your task is to write a SQL query that answers the question. Database file: `/app/db.sqlite` You may use `sqlite3` or Python to inspect the schema and data. Do not modify the database. Question: What is the difference in the annual average consumption of the customers with the least amount of consumption paid in CZK for 2013 between SME and LAM, LAM and KAM, and KAM and SME? Evidence: annual average consumption of customer with the lowest consumption in each segment = total consumption per year / the number of customer with lowest consumption in each segment; Difference in annual average = SME's annual average - LAM's annual average; Difference in annual average = LAM's annual average - KAM's annual average; Year 2013 can be presented as Between 201301 And 201312; The first 4 strings of the Date values in the yearmonth table can represent year. Schema: CREATE TABLE customers ( CustomerID INTEGER UNIQUE not null primary key, Segment TEXT null, Currency TEXT null ) CREATE TABLE gasstations ( GasStationID INTEGER UNIQUE not null primary key, ChainID INTEGER null, Country TEXT null, Segment TEXT null ) CREATE TABLE products ( ProductID INTEGER UNIQUE not null primary key, Description TEXT null ) CREATE TABLE "transactions_1k" ( TransactionID INTEGER primary key autoincrement, Date DATE, Time TEXT, CustomerID INTEGER, CardID INTEGER, GasStationID INTEGER, ProductID INTEGER, Amount INTEGER, Price REAL ) CREATE TABLE "yearmonth" ( CustomerID INTEGER not null references customers on update cascade on delete cascade references customers, Date TEXT not null, Consumption REAL, primary key (Date, CustomerID) ) Output: Write ONLY the SQL query to `/app/answer.sql`. Do not include code fences, comments, or explanations. ``` --- Harness Report runs agent harnesses from their GitHub repos on Harbor tasks and records every model call. Every page is also `.md` and `.json`; index: https://harnessreport.com/llms.txt · MCP: https://harnessreport.com/mcp