Comments
There's unfortunately not much to read here yet...
Best of Machine Learning
Discover the best guides, books, papers and news in Machine Learning, once per week.
Follow the full discussion on Reddit.
I am trying to build a summarizer for a conversation that happened between a rule-based bot and a customer. To my disadvantage the working language is Turkish. I gathered fine-tuning data of 1.000 examples. Also, I have a Turkish summarization dataset of +100k. As far as I observed instruction fine-tuning will yield proper results if and only if there is a good amount of examples in the pre-training data of the LLM. Have you had similar experiences with low-resource languages? Any advice on how to tackle such issues? Also, do you know any open-source LLM with a high amount of low-resource language in its pre-training data?
There's unfortunately not much to read here yet...
Ever having issues keeping up with everything that's going on in Machine Learning? That's where we help. We're sending out a weekly digest, highlighting the Best of Machine Learning.
Best of Machine Learning
Discover the best guides, books, papers and news in Machine Learning, once per week.
Join 800 others in curating the best parts of Machine Learning!
Just one more step. We use social logins to make sure that real humans are curating the list!
or contribute on GitHub directly.