Text Summarization on XSum (ROUGE-2, Rej., Speed)
21.3ROUGE-2VIA-SD
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VIA-SDModel=T5 S→L, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 21.3 | 26 | 2.15 | |
| Faster CascadesModel=T5 S→L, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 19.9 | 30 | 1.42 | |
| Faster CascadesModel=T5 S→XL, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 18.99 | 30 | 1.45 | |
| VIA-SDModel=T5 S→XL, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 18.95 | 28 | 1.65 | |
| Faster CascadesModel=T5 S→XL, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 18.9 | 31 | 1.6 | |
| VIA-SDModel=T5 S→XL, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 18.9 | 26 | 1.95 | |
| Speculative DecodingModel=T5 S→XL, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 18.9 | 38 | 1.12 | |
| BiLDModel=T5 S→XL, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 18.9 | 36 | 1.2 | |
| Cascade SDModel=T5 S→XL, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 18.9 | 35 | 1.25 | |
| SWiFTModel=T5 S→XL, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 18.9 | 34 | 1.3 | |
| CLaSPModel=T5 S→XL, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 18.9 | 33 | 1.35 | |
| Speculative DecodingModel=T5 S→XL, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 18.7 | 36 | 1.28 | |
| BiLDModel=T5 S→XL, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 18.7 | 34 | 1.35 | |
| Cascade SDModel=T5 S→XL, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 18.7 | 33 | 1.4 | |
| SWiFTModel=T5 S→XL, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 18.7 | 32 | 1.45 | |
| CLaSPModel=T5 S→XL, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 18.7 | 31 | 1.48 | |
| Speculative DecodingModel=T5 S→L, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 16.36 | 33 | 1.2 | |
| BiLDModel=T5 S→L, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 16.36 | 32 | 1.25 | |
| Cascade SDModel=T5 S→L, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 16.36 | 31 | 1.3 | |
| SWiFTModel=T5 S→L, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 16.36 | 30 | 1.35 | |
| CLaSPModel=T5 S→L, Decoding Setting=Greedy Decoding, Temperature=02026.06 | 16.36 | 29 | 1.38 | |
| VIA-SDModel=T5 S→L, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 15.27 | 24 | 1.9 | |
| Faster CascadesModel=T5 S→L, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 15.05 | 30 | 1.3 | |
| Speculative DecodingModel=T5 S→L, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 14.8 | 36 | 1.02 | |
| BiLDModel=T5 S→L, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 14.8 | 34 | 1.1 | |
| Cascade SDModel=T5 S→L, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 14.8 | 33 | 1.15 | |
| SWiFTModel=T5 S→L, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 14.8 | 32 | 1.2 | |
| CLaSPModel=T5 S→L, Decoding Setting=Non-Greedy Sampling, Temperature=12026.06 | 14.8 | 31 | 1.25 |