Competitive Intelligence on multi_source_brief Haiku
10,475Input Token Count (Fine)LangGraph (hand-tuned)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| LangGraph (hand-tuned)Worker Model=Claude 3 Haiku, Judge Model=Claude 3 Opus, Runs per cell=152026.05 | 10,475 | 6,897 | 5,572 | 4,018 | 80.7 | 79.8 | — | — | |
| Agent CapsulesWorker Model=Claude 3 Haiku, Judge Model=Claude 3 Opus, Runs per cell=152026.05 | 5,121 | 6,138 | 3,246 | 4,751 | 82.7 | 81.5 | — | — |