Task Execution on TEXTCRAFT-SYNTH Medium (eval)
98Success RateRecursive Agent
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Recursive AgentContext Window (Train/Eval)=40K / 256K2026.05 | 98 | 109 | 20.9 | |
| Single AgentContext Window (Train/Eval)=40K / 256K2026.05 | 87 | 60 | 38.1 |