ResearchBenchmarksFunction Calling on BFCL non-live (test)Follow82AST Accuracy (Simple Python)Ours48.7257.366674.64Jan 25, 2026Evaluation ResultsMethodMethodLinksAST Accuracy (Simple Python)AST Accuracy (Parallel)AST Accuracy (Multiple)AST Accuracy (Parallel Multiple)AST Accuracy (Irrelevance)AST Accuracy (Overall)OursFine-tuning dataset=Pr...Fine-tuning dataset=Proposed Dataset2026.018282918092.985.2ToolAceFine-tuning dataset=To...Fine-tuning dataset=ToolAce2026.0178.377.5795595.477.8APIGenFine-tuning dataset=AP...Fine-tuning dataset=APIGen2026.0161.572.58283.597.577.1UntrainedStatus=Instruction tun...Status=Instruction tuned model baseline without specific fine-tuning2026.015048.55334.51.338.3