Agentic Task Execution on AgentDojo Banking
91.2BURepeat prompt
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Repeat prompt2026.01 | 91.2 | 82.3 | 22.7 | |
| Delimiter2026.01 | 85.6 | 75.2 | 28.3 | |
| AGENTRIM2026.01 | 80 | 72.4 | 0.9 | |
| Baseline2026.01 | 78.7 | 76.5 | 36.5 | |
| Tool filter2026.01 | 70 | 62.1 | 11.4 | |
| PI detector2026.01 | 38.8 | 31.1 | 0.8 |