COLDSELECT jointly selects annotation instances and verbalizer tokens in cold-start prompt-based classification by clustering shared embeddings and greedily minimizing cluster impurity.
In: Proceedings of the 59th Annual Meeting of the ACL and the 11th IJCNLP (V olume 1: Long Papers)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Modeling Data Diversity for Joint Instance and Verbalizer Selection in Cold-Start Scenarios
COLDSELECT jointly selects annotation instances and verbalizer tokens in cold-start prompt-based classification by clustering shared embeddings and greedily minimizing cluster impurity.