List your company

Resource Efficiency on ToolBench (Tool Use)

5.6Energy (Wh)Benign
-120.536730.8821,582.32,433.718Jan 16, 2026

Evaluation results

MethodLinks
BenignTarget Model=Llama-3.32026.015.6289.730
BenignTarget Model=Qwen-32026.016.1240.330
BenignTarget Model=L-DeepSeek2026.016.8334.740
BenignTarget Model=Mistral-L2026.017.3368.140
BenignTarget Model=GLM-4.52026.017.4327.520
OverthinkTarget Model=Llama-3.32026.0111.9331.350
BenignTarget Model=Seed-32B2026.0115.2301.280
OverthinkTarget Model=L-DeepSeek2026.0125.9334.6130
OverthinkTarget Model=Mistral-L2026.0148.3387.1320
OverthinkTarget Model=Seed-32B2026.0148.3311510
OverthinkTarget Model=Qwen-32026.01118.1349.2670
OverthinkTarget Model=GLM-4.52026.01241.6395.3780
MCTS-based tool-layer DoS attackTarget Model=Qwen-32026.011,829.1538.74,400
MCTS-based tool-layer DoS attackTarget Model=GLM-4.52026.012,401447.63,670
MCTS-based tool-layer DoS attackTarget Model=L-DeepSeek2026.012,504.1527.76,560
MCTS-based tool-layer DoS attackTarget Model=Seed-32B2026.012,779.9588.35,170
MCTS-based tool-layer DoS attackTarget Model=Mistral-L2026.013,127.1576.57,390
MCTS-based tool-layer DoS attackTarget Model=Llama-3.32026.013,159572.26,650