Summarization on CNN/DM non-anonymized (test)
100ROUGE-1Human-written Summaries
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Human-written Summariestype=Baseline2021.05 | 100 | 100 | 100 | 39.2 | 61.2 | 58.5 | |
| PEGASUS_LARGEparadigm=supervised, model_size=large2021.05 | 44.1 | 21.3 | 40.9 | — | — | — | |
| PEGASUS_BASEparadigm=supervised, model_size=base2021.05 | 41.8 | 18.8 | 38.9 | — | — | — | |
| Bottom-Upparadigm=supervised2021.05 | 41.2 | 18.7 | 38.3 | 37.8 | 53.8 | 73.9 | |
| Lead-3 baselinetype=Baseline, strategy=extractive2021.05 | 40.3 | 17.7 | 36.6 | 42.1 | 65.6 | 84 | |
| PG + Coverageparadigm=supervised2021.05 | 39.5 | 17.3 | 36.4 | 37.7 | 50.8 | 61.7 | |
| Summary Loop 45paradigm=unsupervised2021.05 | 37.7 | 14.8 | 34.7 | 40.4 | 62.7 | 47 | |
| Pointer Generatorparadigm=supervised2021.05 | 36.4 | 15.7 | 33.4 | 34.2 | 54.7 | 55.6 | |
| TextRankparadigm=unsupervised, strategy=extractive2021.05 | 35.2 | 12.9 | 28.7 | 37 | 61.2 | 49.62 | |
| GPT2 Zero-Shotparadigm=unsupervised, zero-shot=true2021.05 | 29.3 | 8.3 | 26.6 | — | — | — |