Question Answering on TriviaQA 50/50 calibration (test)
74.5CR (%)Optimal pair
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Optimal pair2026.05 | 74.5 | 0.316 | |
| Optimal subsequence2026.05 | 74.5 | 0.306 | |
| Full fixed chain2026.05 | 67.2 | 0.249 | |
| Diagnostic learned router2026.05 | 59.9 | 0.219 | |
| Always-expensive2026.05 | 0 | — |