Seq2seq + selective + MTL + ERAM

Reported on 6 benchmarks across 1 task

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

Knowledge Base6 results

Text SummarizationonDUC 2004 Task 1
ROUGE-1
29.33
best: 33.06 (Transformer+WDrop)
Text SummarizationonDUC 2004 Task 1
ROUGE-2
10.24
best: 17.74 (Transformer LM)
Text SummarizationonDUC 2004 Task 1
ROUGE-L
25.24
best: 28.52 (Transformer+LRPE+PE+Re-ranking+Ensemble)
Text SummarizationonGigaWord
ROUGE-1
35.33
best: 60.12 (OpenAI/o3-mini)
Text SummarizationonGigaWord
ROUGE-2
17.27
best: 54.22 (OpenAI/o3-mini)
Text SummarizationonGigaWord
ROUGE-L
33.19
best: 60.29 (Riple/Saanvi-v0.1)