A concept-bottleneck reward model with Expected Information Gain acquisition learns interpretable concepts faster under limited annotation budgets without hurting preference accuracy.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Interpretable Reward Modeling with Active Concept Bottlenecks
A concept-bottleneck reward model with Expected Information Gain acquisition learns interpretable concepts faster under limited annotation budgets without hurting preference accuracy.