BFCL
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
BFCL out-of-domain
26Exact Match (EM)
10
BFCL v3
75.61Score
10
BFCL Single-Turn
84.11OA
10
BFCL out-of-domain
38.06Exact Match
9
BFCL v3 2025-08-26 (test)
50Multi-Turn Overall Accuracy
9
BFCL Multi-Turn Base v3 (val)
43Avg@8
9
BFCL (test)
90.2Accuracy
9
BFCL Non-Live (Python)
83.8Accuracy
8
BFCL Live Python
69.7Accuracy
8
BFCL Non-Live JSON
85.5Accuracy
8
BFCL Live JSON
71.5Accuracy
8
BFCL
74.41Accuracy
8
BFCL
77Success Rate (SR)
8
BFCL v3
60.5Multi-Turn Success Rate
8
BFCL
79Full Accuracy
8
BFCL All Combined + Irrelevance
78AUROC
8
BFCL Simple + Irrelevance
81AUROC
8
BFCL All Combined
0.76AUROC
8
BFCL Multiple + Parallel-Multiple
77AUROC
8
BFCL Simple + Parallel
75AUROC
8
BFCL Simple + Multiple
77AUROC
8
BFCL Parallel-Multiple
76AUROC
8
BFCL Parallel
0.72AUROC
8
BFCL Multiple
0.76AUROC
8
BFCL Simple
0.77AUROC
8