Human curation in Alchemist puts you in control of your dataset’s composition. You have the freedom to choose which
samples you want to add to your dataset based on your specific requirements and goals. This hands-on approach allows for
fine-grained control over the dataset’s content.
When you come across a particularly good example, Alchemist offers the option to automatically add supplementary or
complementary samples. This feature helps in expanding your dataset with relevant content, saving time and effort in the
curation process.
A key consideration in manual curation is ensuring good representation of different examples in your dataset. This
diversity is crucial for effective instruction fine-tuning, as it helps the model learn from a wide range of scenarios
and inputs. Alchemist’s interface is designed to support this goal, making it easy to review and select a variety of samples.
For those looking to deepen their understanding of effective curation strategies, Alchemist’s blog offers in-depth resources.