spacemanidol

/

flan-t5-large-xsum

like 2

Hello, I am using the template -"Summarize:\n\n{text}", "{summary}", num_beams=3 and sampling turned on while evaluating the test split. The ROUGE-1 score is 0.383 which is not an improvement over the non-finetuned version.

Could you reveal your setup? Thanks.