ConceptHierarchical Loss
Hierarchical Loss
A loss function introduced in the HASTE framework that computes supervision signals separately at each prefix level of the semantic ID, explicitly returning the tree-structured hierarchical prior of semantic IDs to the model and thereby accelerating training convergence.
Timeline (last 90 days)
Oct 14
HASTE's hierarchical loss provides a supervision signal at every level of the item hierarchy, making the gradient signal denser and more directional, thereby significantly accelerating training convergence.
Unverified50%