News Recommendation on Offline News Recommendation Dataset
25.7Recall@5Ours (RL-trained)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Ours (RL-trained)Training Strategy=RL-trained, Backbone=Qwen2.5, Model Size=32B2026.02 | 25.7 | 27.1 | 20.9 | 24.1 | 19.9 | |
| Ours (Distilled)Training Strategy=Distilled, Backbone=Qwen2.5, Model Size=0.5B2026.02 | 23.3 | 26.1 | 19.6 | 23.5 | 18 | |
| HSTU2026.02 | 21.6 | 25 | 19.4 | 22.8 | 18.4 | |
| PinSage2026.02 | 16.8 | 22 | 14.9 | 18.4 | 13.8 | |
| NRHUB2026.02 | 16.2 | 21.2 | 14.5 | 17.8 | 13.3 | |
| SASRec2026.02 | 15.8 | 20.6 | 14.1 | 17.3 | 12.9 | |
| GRU2026.02 | 14.4 | 18.9 | 12.9 | 15.9 | 11.7 |