Go back
Model
arXiv Abstract-to-Title · mT5
Wetin e dey do
Give di model one research abstract and e go suggest title for di paper. E learn dis work as dem fine-tune mT5-small with pairs of abstracts and their titles from arXiv. Each input fit take up to 256 tokens during training. Di ROUGE scores wey dem publish dey measure words or word sequences wey match di original title. Good title wey use different words fit still score low.
Original documents
Original source · huggingface.co mT5-smallHow to cite am
arXiv Abstract-to-Title · mT5 · https://huggingface.co/emre/arxiv27k-t5-abst-title-gen