Single-agent tool use on API-Bank reconstructed
79.27CorrectnessNaive Agent
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Naive AgentTool Num=22026.01 | 79.27 | — | 10.43 | |
| SEAgentTool Num=12026.01 | 74.73 | 0 | 9 | |
| SEAgentTool Num=22026.01 | 71.34 | 0 | 11.16 | |
| Naive AgentTool Num=12026.01 | 70.33 | — | 10.8 | |
| SEAgentTool Num=>= 32026.01 | 67.91 | 5.13 | 12.79 | |
| Naive AgentTool Num=>= 32026.01 | 63.43 | — | 15.54 | |
| IsolateGPTTool Num=12026.01 | 53.95 | 5.26 | 21.98 | |
| IsolateGPTTool Num=22026.01 | 37.32 | 18.31 | 34.08 | |
| IsolateGPTTool Num=>= 32026.01 | 37.25 | 5.88 | 64.42 |