List your company
ResearchFormerly WizWand

Tool Use on RobustBench-TC Clean 1.0 (test)

77.9Clean AccuracyLoopTool-32B
50.75657.80364.8571.897May 12, 2026
Updated 5mo ago

Evaluation results

MethodLinks
LoopTool-32BModel Category=Open-source RL tool models, Model Scale=32B2026.0577.9
Qwen3-32BModel Category=Base / instruct counterparts, Model Scale=32B2026.0575.4
Qwen3-14BModel Category=Base / instruct counterparts, Model Scale=14B2026.0574.9
LoopTool-8BModel Category=Open-source RL tool models, Model Scale=8B2026.0571.4
Qwen3.5-9BModel Category=Frontier, Model Scale=9B2026.0570.9
o4-mini (OpenAI)Model Category=Closed-source frontier2026.0570.9
ToolRL-Qwen2.5-1.5BModel Category=Open-source RL tool models, Model Scale=1.5B2026.0570.4
MUA-RL-32BModel Category=Open-source RL tool models, Model Scale=32B2026.0568.3
MUA-RL-8BModel Category=Open-source RL tool models, Model Scale=8B2026.0567.8
Qwen3-8BModel Category=Base / instruct counterparts, Model Scale=8B2026.0567.3
ToolRL-DR-Mixed (ours)Model Category=Our 3B trained checkpoints, Model Scale=3B2026.0565.8
TL-CodeLLaMA-2Model Category=Open-source RL tool models, Model Scale=7B2026.0565.3
ToolRL-DR-Full (ours)Model Category=Our 3B trained checkpoints, Model Scale=3B2026.0564.3
ToolRL-Qwen2.5-3BModel Category=Open-source RL tool models, Model Scale=3B2026.0563.8
MUA-RL-14BModel Category=Open-source RL tool models, Model Scale=14B2026.0562.8
ToolRL-Llama3.2-3BModel Category=Open-source RL tool models, Model Scale=3B2026.0562.8
DeepSeek-R1-Distill-14BModel Category=Frontier, Model Scale=14B2026.0561.8
Qwen2.5-3B-InstructModel Category=Base / instruct counterparts, Model Scale=3B2026.0560.8
SFT-Clean-4k-3BModel Category=ToolRL SFT version, Model Scale=3B2026.0559.8
Qwen2.5-1.5B-InstructModel Category=Base / instruct counterparts, Model Scale=1.5B2026.0557.3
Llama-3.2-3B-InstructModel Category=Base / instruct counterparts, Model Scale=3B2026.0551.8