Text Summarization on XSUM (AlignScore, BARTScore, FactCC, ROUGE-L)
88.7AlignScoreMultiMetric
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MultiMetricbackbone=LLaMA2026.05 | 88.7 | -3.47 | 77.4 | 18.3 | |
| MultiMetricbackbone=BART2026.05 | 86.6 | -3.63 | 83.3 | 33.9 | |
| SFTbackbone=LLaMA2026.05 | 86.1 | -3.48 | 75.1 | 19.2 | |
| MultiMetricbackbone=DeepSeek2026.05 | 83.2 | -3.23 | 80.5 | 14 | |
| SFTbackbone=DeepSeek2026.05 | 82.5 | -3.34 | 75.9 | 14.8 | |
| MPObackbone=DeepSeek2026.05 | 81.3 | -3.39 | 78.8 | 12.5 | |
| MPObackbone=LLaMA2026.05 | 79.8 | -3.49 | 76 | 18.8 | |
| MultiMetricbackbone=GPT-J2026.05 | 75.8 | -3.57 | 80.6 | 22.3 | |
| MPObackbone=BART2026.05 | 62 | -3.67 | 76 | 33.5 | |
| SFTbackbone=BART2026.05 | 61.9 | -3.69 | 65.7 | 36.4 | |
| SFTbackbone=GPT-J2026.05 | 59.7 | -3.63 | 62.4 | 25 | |
| MPObackbone=GPT-J2026.05 | 53.5 | -3.9 | 73.8 | 23.6 |