|
--- |
|
language: |
|
- en |
|
tags: |
|
- text2text-generation |
|
- pytorch |
|
license: "gpl-3.0" |
|
datasets: |
|
- xsum |
|
widget: |
|
- text: "President Biden met with Russia's Putin over the weekend to discuss a ceasefire in Ukraine." |
|
example_title: "Ukrainian Ceasefire" |
|
- text: "Acme Ventures recently led a seed round to provide over $2MM in funding to Aiko Mail, an AI startup tackling email." |
|
example_title: "VC Investment" |
|
- text: "In a shocking move, Florida has decided to formally secede from the United States, opting to sink into the Atlantic Ocean." |
|
example_title: "Florida secedes" |
|
--- |
|
|
|
# Maeve - XSUM |
|
|
|
Maeve is a language model that is similar to BART in structure but trained specially using a CAT (Conditionally Adversarial Transformer). |
|
|
|
This allows the model to learn to create long-form text from short entries with high degrees of control and coherence that are impossible to achieve with traditional transformers. |
|
|
|
This specific model has been trained on the XSUM dataset, and can invert summaries into full-length news articles. Feel free to try examples on the right! |
|
|