The authors extend behavioral entropy to continuous spaces, derive k-nearest-neighbor estimators and a reward function, and show that datasets generated by this objective improve downstream offline RL performance over standard entropy-based baselines.
Initial trials showed q ∈ {2.0, 3.0, 5.0} led to performance no better (and usually worse) than q = 1 .1, so offline RL training for these q values was not performed
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
-
Behavioral Entropy-Guided Dataset Generation for Offline Reinforcement Learning
The authors extend behavioral entropy to continuous spaces, derive k-nearest-neighbor estimators and a reward function, and show that datasets generated by this objective improve downstream offline RL performance over standard entropy-based baselines.