Scheduled Commit

data/evaluation_fc95fde5-ac55-4a9c-84f2-e9e082c4f013.json

@@ -1 +1,2 @@
 {"inference_api": "openrouter", "model_name": "qwen/qwen-2.5-coder-32b-instruct", "prompt_format": "duckdbinstgraniteshort", "custom_prompt": "", "timestamp": "2024-11-12T04:18:20.511844", "easy_count": 5, "easy_execution_accuracy": 0.8, "medium_count": 14, "medium_execution_accuracy": 0.8571428571428571, "hard_count": 6, "hard_execution_accuracy": 0.16666666666666666, "duckdb_count": 48, "duckdb_execution_accuracy": 0.6041666666666666, "ddl_count": 2, "ddl_execution_accuracy": 1.0, "all_count": 75, "all_execution_accuracy": 0.64}
+{"inference_api": "openrouter", "model_name": "qwen/qwen-2.5-coder-32b-instruct", "prompt_format": "custom_85307618", "custom_prompt": "Generate a valid DuckDB SQL query to accurately answer the user's question. Only respond with the SQL code.\n\nDuckDB SQL Specifics:\n\nIdentifiers and Literals: Use double quotes (\") for identifiers (tables, columns) with spaces or special characters, and single quotes (') for string literals.\nDirect File Querying: DuckDB supports querying directly from CSV, Parquet, and JSON files, e.g., SELECT * FROM 'data.csv';\nCTAS: Supports CREATE TABLE AS (CTAS) syntax, e.g., CREATE TABLE new_table AS SELECT * FROM old_table;\nFROM Clauses: Queries can start with FROM, omitting SELECT * (e.g., FROM my_table WHERE condition;).\nExpression Evaluation: Use SELECT without FROM to generate expressions or single results, e.g., SELECT 1 + 1;.\nMultiple Database Support: Attach databases using ATTACH and refer with dot notation, e.g., ATTACH 'db.duckdb' AS mydb; then mydb.table_name.\nImplicit Type Conversion: DuckDB allows implicit type conversions, or make them explicit with :: (e.g., '42'::INTEGER + 1;).\nString and Array Slicing: Use [start:end] or [start:end:step] for slicing (strings, arrays), with 1-based indexes.\nPatterned Column Selection: Select columns with patterns, e.g., SELECT COLUMNS('sales_.*') FROM sales_data;.\nExclusion/Inclusion with Wildcards: Exclude/include or transform columns with SELECT * EXCLUDE (col); and SELECT * REPLACE (col AS new_col);.\nGrouping/Ordering: GROUP BY ALL or ORDER BY ALL to group/order by non-aggregated columns.\nUnion by Column Name: Combine tables by matching column names, e.g., table1 UNION BY NAME table2;.\nComplex Types: Support for List, Struct, Map, and Array types with intuitive syntax ([1, 2, 3] AS my_list;).\nField Access: Access struct fields with dot notation, and maps with brackets.\nDate and Timestamp Conversion: strftime(NOW(), '%Y-%m-%d'); for the current date, and strptime('YYYY-MM-DD', '%Y-%m-%d')::TIMESTAMP; for string-to-timestamp.\nAlias Usage: Column aliases in the WHERE, GROUP BY, and HAVING clauses, e.g., SELECT a + b AS total FROM my_table WHERE total > 10 GROUP BY total;.\nList Comprehensions: Generate lists with expressions, e.g., SELECT [x*2 FOR x IN [1, 2, 3]];.\nFunction Chaining: Chain functions with . (dot operator), e.g., SELECT 'DuckDB'.replace('Duck', 'Goose');.\nJSON Handling: Use -> and ->> for JSON fields, e.g., data->'$.user.id' AS user_id;.\nRegex Functions: regexp_matches(), regexp_replace(), and regexp_extract() are supported.\nData Sampling: Retrieve a data subset with USING SAMPLE, e.g., SELECT * FROM large_table USING SAMPLE 10%;.\nDatabase Schema: {schema}\n\nQuestion: {question}\n```\n", "timestamp": "2024-11-12T04:40:38.781499", "easy_count": 5, "easy_execution_accuracy": 0.6, "medium_count": 14, "medium_execution_accuracy": 0.8571428571428571, "hard_count": 6, "hard_execution_accuracy": 0.16666666666666666, "duckdb_count": 48, "duckdb_execution_accuracy": 0.5833333333333334, "ddl_count": 2, "ddl_execution_accuracy": 1.0, "all_count": 75, "all_execution_accuracy": 0.6133333333333333}