Share your thoughts, 1 month free Claude Pro on usSee more

Tool Use on RobustBench-TC Clean 1.0 (test)

77.9Clean Accuracy

LoopTool-32B

Updated 21d ago

Evaluation Results

Method	Links
LoopTool-32B 2026.05		77.9
Qwen3-32B 2026.05		75.4
Qwen3-14B 2026.05		74.9
LoopTool-8B 2026.05		71.4
Qwen3.5-9B 2026.05		70.9
o4-mini (OpenAI) 2026.05		70.9
ToolRL-Qwen2.5-1.5B 2026.05		70.4
MUA-RL-32B 2026.05		68.3
MUA-RL-8B 2026.05		67.8
Qwen3-8B 2026.05		67.3
ToolRL-DR-Mixed (ours) 2026.05		65.8
TL-CodeLLaMA-2 2026.05		65.3
ToolRL-DR-Full (ours) 2026.05		64.3
ToolRL-Qwen2.5-3B 2026.05		63.8
MUA-RL-14B 2026.05		62.8
ToolRL-Llama3.2-3B 2026.05		62.8
DeepSeek-R1-Distill-14B 2026.05		61.8
Qwen2.5-3B-Instruct 2026.05		60.8
SFT-Clean-4k-3B 2026.05		59.8
Qwen2.5-1.5B-Instruct 2026.05		57.3
Llama-3.2-3B-Instruct 2026.05		51.8