Expectile n-step Q-learning (ENQ) applies an upper-expectile loss to the action-value TD error, reducing the pessimistic bias of multi-step returns in off-policy RL; the authors prove contraction and bias bounds and report competitive results across 27 tasks.
Proceedings of the AAAI conference on artificial intelligence , volume=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Upper-Expectile Multi-Step Q-Learning for Off-Policy Reinforcement Learning
Expectile n-step Q-learning (ENQ) applies an upper-expectile loss to the action-value TD error, reducing the pessimistic bias of multi-step returns in off-policy RL; the authors prove contraction and bias bounds and report competitive results across 27 tasks.