PB² pairs a population of agents with a discriminative diversity bonus to improve query distinguishability, feedback efficiency, and robustness to noisy human preferences in PbRL.
Autonomous skill discovery with quality-diversity and unsupervised descriptors
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
PB$^2$: Preference Space Exploration via Population-Based Methods in Preference-Based Reinforcement Learning
PB² pairs a population of agents with a discriminative diversity bonus to improve query distinguishability, feedback efficiency, and robustness to noisy human preferences in PbRL.