List your company

Tool Calling on API-Bank L-2

90.42Name Match F1HiTEC-KTO
50.785661.075371.36581.6547May 28, 2025

Evaluation results

MethodLinks
HiTEC-KTOModel=Llama3-8B2025.0590.42——64.2
+ Function CallingModel=GPT-3.5-Turbo-01252025.0588.72——65.87
HiTEC-KTOModel=Qwen2.5-7B2025.0588——59.93
HiTEC-KTOModel=Llama3.1-8B2025.0585.61——61.02
HiTEC-ICLModel=GPT-3.5-Turbo-01252025.0583.8——64.72
HiTEC-ICLModel=Llama3.1-8B2025.0582.23——51.29
HiTEC-ICLBackbone=Llama3.1-8B2025.0582.23——51.29
VanillaModel=GPT-3.5-Turbo-01252025.0578.87——61.02
+ Function CallingModel=GPT-4-Turbo2025.0578.49——62.32
Function CallingBackbone=GPT-4-Turbo2025.0578.49——62.32
VanillaModel=Llama3.1-8B2025.0578.23——50.25
VanillaBackbone=Llama3.1-8B2025.0578.23——50.25
HiTEC-ICLModel=Llama3-8B2025.0577.3——56.01
HiTEC-ICLBackbone=Llama3-8B2025.0577.3——56.01
VanillaModel=Llama3-8B2025.0575.54——48.91
VanillaBackbone=Llama3-8B2025.0575.54——48.91
CoTBackbone=Llama3.1-8B2025.0575.54——48.52
HiTEC-ICLModel=GPT-4-Turbo2025.0571.21——72.68
HiTEC-ICLBackbone=GPT-4-Turbo2025.0571.21——72.68
VanillaModel=GPT-4-Turbo2025.0571.09——70.43
VanillaBackbone=GPT-4-Turbo2025.0571.09——70.43
+ CoTModel=GPT-4-Turbo2025.0567.86——65.74
CoTBackbone=GPT-4-Turbo2025.0567.86——65.74
+ CoTModel=GPT-3.5-Turbo-01252025.0559.2——55.53
CoTBackbone=Llama3-8B2025.0552.31——45.65
Base RetrieverBase Model=Qwen3-30B2026.01—57.8——
Hammer2-0.5BModel Series=Qwen2.5 Series, Method=Baseline2025.05—43.3238.22—
Hammer2-1.5BModel Series=Qwen2.5 Series, Method=Baseline2025.05—80.5162.82—
Hammer2-3BModel Series=Qwen2.5 Series, Method=Baseline2025.05—77.1157.58—
Hammer2-7BModel Series=Qwen2.5 Series, Method=Baseline2025.05—75.9658.36—
Llama3-70BModel Series=Llama3 Series, Method=Baseline2025.05—66.6763.65—
Llama3-70BModel Series=Llama3 Series, Method=HiTEC-KTO2025.05—75.6163.08—
Llama3-70BModel Series=Llama3.1 Series, Method=HiTEC-KTO2025.05—86.4465.14—
Llama3-8BModel Series=Llama3 Series, Method=Baseline2025.05—75.5448.91—
Llama3-8BModel Series=Llama3 Series, Method=HiTEC-KTO2025.05—90.4264.2—
Llama3.1-70BModel Series=Llama3.1 Series, Method=Baseline2025.05—80.3462.2—
Llama3.1-8BModel Series=Llama3.1 Series, Method=Baseline2025.05—78.2350.25—
Llama3.1-8BModel Series=Llama3.1 Series, Method=HiTEC-KTO2025.05—85.6161.02—
Q2DBase Model=Qwen3-30B2026.01—58.5——
Q2EBase Model=Qwen3-30B2026.01—59.5——
Q2PBase Model=Qwen3-30B2026.01—56.1——
Qwen2.5-0.5BModel Series=Qwen2.5 Series, Method=HiTEC-KTO2025.05—81.7654.52—
Qwen2.5-1.5BModel Series=Qwen2.5 Series, Method=HiTEC-KTO2025.05—81.3660.66—
Qwen2.5-3BModel Series=Qwen2.5 Series, Method=HiTEC-KTO2025.05—85.5154.61—
Qwen2.5-7BModel Series=Qwen2.5 Series, Method=HiTEC-KTO2025.05—8859.93—
Re-InvokeBase Model=Qwen3-30B2026.01—60.5——
TOOLQPBase Model=Qwen3-30B2026.01—60.7——