Question Answering on Dataset A v0.4 (test)
4,582Solved Countthree-model oracle
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| three-model oraclemodel_pool=qwen, ds4, kimi2026.06 | 4,582 | 83.25 | |
| kimirouting_strategy=always2026.06 | 4,129 | 75.02 | |
| ds4routing_strategy=always2026.06 | 4,056 | 73.69 | |
| qwenrouting_strategy=always2026.06 | 3,477 | 63.17 | |
| unsolvable (any model)model_pool=qwen, ds4, kimi2026.06 | 922 | 16.75 |