Summarization on XSum (LLM/GEval Metrics)
0.761MiniCheck ScoreSCRPO
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SCRPOvariant=Full2025.12 | 0.761 | 4.38 | 4.12 | 4.66 | 2.99 | 4.23 | |
| SCRPOvariant=Critique only2025.12 | 0.738 | 4.33 | 4.11 | 4.62 | 2.99 | 4.24 | |
| SCRPOvariant=SFT2025.12 | 0.735 | 4.23 | 4.04 | 4.49 | 2.99 | 4.18 | |
| SCRPOvariant=Inference time2025.12 | 0.722 | 4.23 | 4.06 | 4.44 | 2.99 | 4.16 | |
| SCOPE2025.12 | 0.713 | 4.14 | 4.06 | 4.42 | 2.99 | 4.18 | |
| Pretrained LLM2025.12 | 0.701 | 4.16 | 4.04 | 4.43 | 2.99 | 4.18 | |
| MPO2025.12 | 0.694 | 4.13 | 4.03 | 4.4 | 2.98 | 4.17 |