Text Summarization on CNN/DailyMail LLMs-9 (test)
44.8ROUGE-1SCURank
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SCURanktraining=Distilled ranking method, evaluation=mean ± std over 5 runs2026.04 | 44.8 | 20.5 | 30.6 | 51.7 | 70 | -2.34 | |
| BLANCtraining=Distilled ranking method, evaluation=mean ± std over 5 runs2026.04 | 44.1 | 20.3 | 30.2 | 50.1 | 69.4 | -2.34 | |
| BERTScoretraining=Distilled ranking method, evaluation=mean ± std over 5 runs2026.04 | 43.9 | 20 | 30.1 | 52.2 | 69.5 | -2.46 | |
| UnRanktraining=MLE-only baseline, evaluation=mean ± std over 5 runs2026.04 | 43.8 | 20.5 | 30.1 | 53.3 | 69.9 | -2.59 | |
| GPTRanktraining=Distilled ranking method, evaluation=mean ± std over 5 runs2026.04 | 43.8 | 20 | 30.1 | 51.4 | 69.7 | -2.36 | |
| ROUGEtraining=Distilled ranking method, evaluation=mean ± std over 5 runs2026.04 | 43.2 | 20.4 | 29.9 | 51.4 | 69 | -2.58 |