Re-ranking on Amazon Sports (test)
0.2461Recall@1RL-zeroshot
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RL-zeroshotTraining Strategy=RL, Sampling Strategy=zeroshot2026.02 | 0.2461 | 0.7012 | 0.9447 | 0.4772 | 0.5732 | |
| RL-targeted-KPTraining Strategy=RL, Sampling Strategy=targeted, Keyphrase setting=KP2026.02 | 0.2428 | 0.7083 | 0.9583 | 0.4802 | 0.5745 | |
| Pre-rank (STL)Training Strategy=STL (mid-training with single-task)2026.02 | 0.2422 | 0.7077 | 0.9583 | 0.4796 | 0.5742 | |
| RL-rejection-KPTraining Strategy=RL, Sampling Strategy=rejection, Keyphrase setting=KP2026.02 | 0.2415 | 0.7077 | 0.9583 | 0.4791 | 0.5737 | |
| RL-targeted-noKPTraining Strategy=RL, Sampling Strategy=targeted, Keyphrase setting=noKP2026.02 | 0.2383 | 0.7083 | 0.9583 | 0.4786 | 0.573 | |
| RL-rejection-noKPTraining Strategy=RL, Sampling Strategy=rejection, Keyphrase setting=noKP2026.02 | 0.237 | 0.7044 | 0.9583 | 0.4743 | 0.5701 | |
| SFT-rejection-KPTraining Strategy=SFT, Sampling Strategy=rejection, Keyphrase setting=KP2026.02 | 0.2363 | 0.709 | 0.9525 | 0.4723 | 0.5664 | |
| SFT-targeted-noKPTraining Strategy=SFT, Sampling Strategy=targeted, Keyphrase setting=noKP2026.02 | 0.2279 | 0.7103 | 0.9596 | 0.475 | 0.5688 | |
| SFT-rejection-noKPTraining Strategy=SFT, Sampling Strategy=rejection, Keyphrase setting=noKP2026.02 | 0.2233 | 0.6608 | 0.9427 | 0.4437 | 0.5532 | |
| SFT-targeted-KPTraining Strategy=SFT, Sampling Strategy=targeted, Keyphrase setting=KP2026.02 | 0.179 | 0.7025 | 0.9557 | 0.4491 | 0.5454 |