Local Joint Self-attention

Reported on 3 benchmarks across 1 task · 1 paper · 1 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Natural Language Processing3 results

Machine TranslationonIWSLT2014 German-English
BLEU score· 2019-05-16
35.7
best: 40.43 (PiNMT)
SOTA
Joint Source-Target Self Attention with Locality Constraints arXiv:1905.06596
Machine TranslationonWMT2014 English-German
BLEU score· 2019-05-16
29.7
best: 35.14 (Transformer Cycle (Rev))
Joint Source-Target Self Attention with Locality Constraints arXiv:1905.06596
Machine TranslationonWMT2014 English-French
BLEU score· 2019-05-16
43.3
best: 46.4 (Transformer+BT (ADMIN init))
Joint Source-Target Self Attention with Locality Constraints arXiv:1905.06596