Then, we try to zero-shot these fine-tuned models on other language pairs. Some are related to French-Catalan, such as Occitan, and some are entirely unrelated, such as Mandarin.

We see it does transfer to Occitan-French pairs and in Cantonese-Mandarin pairs (more on the paper)

We fine-tuned them on Catalan queries and French Docs and see that we can regularise the models to be more robust on Catalan (and we see some gains in French!)

🚨 New Pre-Print!🚨 with @macavaney.bsky.social &
@iadhounis.bsky.social. Stop using "translate-train" for all your multilingual needs. We explore zero-shot transfer for low-resource languages... 🧵