Machine Translation (X→En) on TED (test)
0.22BLEU (Low Resource)many-to-one + xsim
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| many-to-one + xsimModel Type=many-to-one, Training Strategy=Auxiliary Similarity Loss (xsim)2026.01 | 0.22 | 0.288 | 0.266 | |
| many-to-many + xsimModel Type=many-to-many, Training Strategy=Auxiliary Similarity Loss (xsim)2026.01 | 0.218 | 0.307 | 0.304 | |
| many-to-many + multi-parallelModel Type=many-to-many, Training Strategy=multi-parallel batches2026.01 | 0.208 | 0.301 | 0.31 | |
| many-to-manyModel Type=many-to-many, Training Strategy=baseline2026.01 | 0.205 | 0.302 | 0.312 | |
| many-to-oneModel Type=many-to-one, Training Strategy=baseline2026.01 | 0.199 | 0.279 | 0.272 | |
| many-to-one + multi-parallelModel Type=many-to-one, Training Strategy=multi-parallel batches2026.01 | 0.197 | 0.276 | 0.27 |