TARA is a dataset for tool-augmented reward modeling, which includes comprehensive comparison data of human preferences and detailed tool invocation processes.
TARA comprises a total of 13,604 training datasets and 1,469 test sets, each consisting of a question, a positive answer, and a negative answer. TARA incorporates a diverse set of seven tools that span across various domains, encompassing mathematical operations, code-related inquiries, closed-ended and open-ended question answering, knowledge-based queries, and time-sensitive information requests.
Refer to github for details.
Paper | Code | Results | Date | Stars |
---|