Agent Tool-Use on WorkBench
92.59Analytics Success RateMemToolAgent
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MemToolAgentBackbone=Claude 4 Sonnet2026.06 | 92.59 | 93.94 | 84.72 | 88.89 | 68.78 | 83.06 | |
| Extended ThinkingBackbone=Claude 4 Sonnet2026.06 | 55.56 | 88.89 | 76.39 | 88.89 | 57.14 | 69.76 | |
| No MemoryBackbone=Claude 4 Sonnet2026.06 | 50 | 91.92 | 68.06 | 87.65 | 57.14 | 67.94 | |
| ReActBackbone=Claude 4 Sonnet2026.06 | 48.15 | 90.91 | 79.17 | 88.89 | 58.2 | 69.4 |