Task Execution on TEXTCRAFT-SYNTH All (eval)
96Success RateRecursive Agent
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Recursive AgentContext Window (Train/Eval)=40K / 256K2026.05 | 96 | 115 | 19.8 | |
| Single AgentContext Window (Train/Eval)=40K / 256K2026.05 | 73 | 54 | 35.7 |
Loading the SOTA2 catalog…