Binary rewards make the set of reward-maximizing policies infinite in policy gradients; KL control selects the filtered base model but misspecification drives collapse to concentrated valid outputs instead.
Title resolution pending
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.LG 2years
2026 2representative citing papers
One KL-difference identity plus non-negativity of KL derives convexity of the log-partition function, Gibbs variational principle, Pythagorean theorems, and tilting formulas for exponential families.
citing papers explorer
-
Binary Rewards and Reinforcement Learning: Fundamental Challenges
Binary rewards make the set of reward-maximizing policies infinite in policy gradients; KL control selects the filtered base model but misspecification drives collapse to concentrated valid outputs instead.
-
Exponential families from a single KL identity
One KL-difference identity plus non-negativity of KL derives convexity of the log-partition function, Gibbs variational principle, Pythagorean theorems, and tilting formulas for exponential families.