Dialogue Generation on LQA (test)
0.0927BLEU-1Top-p (DDS)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Top-p (DDS)Decoding Strategy=Top-p, DDS=true, Zero-shot=true, Backbone=EVA2.02024.06 | 0.0927 | 0.0558 | 0.0337 | 0.0203 | 0.1172 | 0.1006 | 0.013 | 0.1024 | |
| Typical (DDS)Decoding Strategy=Typical, DDS=true, Zero-shot=true, Backbone=EVA2.02024.06 | 0.0923 | 0.0555 | 0.0336 | 0.0202 | 0.1106 | 0.0931 | 0.0116 | 0.0958 | |
| Top-k (DDS)Decoding Strategy=Top-k, DDS=true, Zero-shot=true, Backbone=EVA2.02024.06 | 0.0921 | 0.0557 | 0.0339 | 0.0206 | 0.1181 | 0.101 | 0.0136 | 0.1043 | |
| Top-p (fixed T)Decoding Strategy=Top-p, DDS=false, Zero-shot=true, Backbone=EVA2.02024.06 | 0.0844 | 0.0509 | 0.0309 | 0.0187 | 0.1143 | 0.099 | 0.0115 | 0.0984 | |
| Top-k (fixed T)Decoding Strategy=Top-k, DDS=false, Zero-shot=true, Backbone=EVA2.02024.06 | 0.0823 | 0.0495 | 0.0299 | 0.018 | 0.1139 | 0.0983 | 0.0113 | 0.0988 | |
| Temperature (DDS)Decoding Strategy=Temperature, DDS=true, Zero-shot=true, Backbone=EVA2.02024.06 | 0.0801 | 0.0482 | 0.0292 | 0.0177 | 0.1041 | 0.0896 | 0.0115 | 0.0918 | |
| Temperature (fixed T)Decoding Strategy=Temperature, DDS=false, Zero-shot=true, Backbone=EVA2.02024.06 | 0.0762 | 0.0452 | 0.0271 | 0.0162 | 0.0656 | 0.0586 | 0.0028 | 0.0568 | |
| Typical (fixed T)Decoding Strategy=Typical, DDS=false, Zero-shot=true, Backbone=EVA2.02024.06 | 0.0554 | 0.0331 | 0.02 | 0.012 | 0.0853 | 0.0724 | 0.0049 | 0.0743 |