The meaning of the suffixes in the folder names are listed as follows: SuffixDescription testOur test data with retrieved contexts from knowledge base v3Contexts in the data are from sentence retrieval v4Contexts in the data are from iterative entity retrieval ...
we improve the evaluation metric to better assess function call correctness by splitting keyword arguments into two groups. We use exact match for keyword arguments that have a unique ground truth, and ROUGE score for those that accept any string with the same semantic meaning as the reference va...