Generative Recommendation on e-commerce dataset (test)
2.46HR@1AdaGRPO
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| AdaGRPOEvaluation state=final training checkpoint, Clipping condition=f1 & f22026.06 | 2.46 | 11.63 | 25.43 | 15.08 | 23.31 | 26.98 | 29.5 | 46.17 | 54.87 | 27 | |
| AdaGRPOEvaluation state=final training checkpoint, Clipping condition=f12026.06 | 2.4 | 11.48 | 25.12 | 15.21 | 23.45 | 27.1 | 29.58 | 45.89 | 54.51 | 31 | |
| BaseEvaluation state=final training checkpoint2026.06 | 2.31 | 11.01 | 24.44 | 12.32 | 19.81 | 24.03 | 25.75 | 42.47 | 52.12 | 8 | |
| GRPO+NLLEvaluation state=final training checkpoint2026.06 | 2.3 | 11.06 | 24.14 | 16.29 | 23.91 | 27.47 | 29.76 | 44.79 | 52.8 | 59 | |
| GRPOEvaluation state=final training checkpoint2026.06 | 2.2 | 10.39 | 22.49 | 14.02 | 21.58 | 25.56 | 29.15 | 44.5 | 52.6 | 89 |