| ID | Benchmark (bleu) | Output | OPUS-MT | bleu | external | bleu | Diff |
|---|---|---|---|---|---|---|---|
| 0 | flores101-devtest | compare | zlw-zle/opus...2022-05-08 | 24.8 | facebook/m2m100_1.2B | 22.8 | 2.0 |
| 1 | flores200-devtest | compare | zlw-zle/opus...2022-05-08 | 24.8 | facebook/m2m100_1.2B | 22.8 | 2.0 |
| 2 | newstest2012 | compare | zlw-zle/opus...2022-05-08 | 21.3 | facebook/nll...illed-1.3B | 22.2 | -0.9 |
| 3 | newstest2013 | compare | zlw-zle/opus...2022-05-08 | 27.9 | facebook/nll...illed-1.3B | 26.6 | 1.3 |
| 4 | ntrex128 | compare | zlw-zle/opus...2022-05-08 | 23.7 | facebook/nll...illed-1.3B | 22.1 | 1.6 |
| 5 | tatoeba-test-v2020-07-28 | compare | zlw-zle/opus...2022-03-19 | 56.5 | facebook/nll...illed-1.3B | 53.7 | 2.8 |
| 6 | tatoeba-test-v2021-03-30 | compare | zlw-zle/opus...2022-03-23 | 56.4 | facebook/nll...illed-1.3B | 53.4 | 3.0 |
| 7 | tatoeba-test-v2021-08-07 | compare | zlw-zle/opus...2022-03-23 | 56.5 | facebook/nll...illed-1.3B | 53.5 | 3.0 |
| average | 36.5 | 34.6 | 1.8 |