Knowledge Graph navigation on AgentBench KG
72.7Task Completion Score (TCS)GPT-5-mini
Evaluation Results
| Method | Links | |
|---|---|---|
| GPT-5-miniAverage Steps (Stp)=9.3, Input Tokens (Tok In)=29.6, Output Tokens (Tok Out)=7.3, Execution Time=89.42026.05 | 72.7 | |
| Gemini-3-flashAverage Steps (Stp)=9.2, Input Tokens (Tok In)=52.9, Output Tokens (Tok Out)=1.1, Execution Time=77.62026.05 | 44 | |
| DeepSeek-V3.2Average Steps (Stp)=20.0, Input Tokens (Tok In)=84.9, Output Tokens (Tok Out)=1.4, Execution Time=69.92026.05 | 34 | |
| Qwen3-235B-A22BAverage Steps (Stp)=11, Input Tokens (Tok In)=43.7, Output Tokens (Tok Out)=0.5, Execution Time=67.22026.05 | 31.3 |