Synthetic DPO preference pairs for training LLMs to use tools correctly in agentic settings.
web_search — real-time web search
calculator — mathematical expression evaluation
weather_api — current weather retrieval
code_interpreter — Python code execution
database_query — SQL database queries
stock_price — financial data lookup
translate — multilingual… See the full description on the dataset page:
https://huggingface.co/datasets/stindardlogic/agentic-dpo-tool-use-3k.