RLEASE is a reinforcement-learning method that learns a neural-network policy to select compact, geometry-dependent active spaces for multireference calculations, trained on a small set of molecules and shown to transfer without retraining.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
physics.chem-ph 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
RLEASE: Reinforcement Learning Efficient Active Space Engine
RLEASE is a reinforcement-learning method that learns a neural-network policy to select compact, geometry-dependent active spaces for multireference calculations, trained on a small set of molecules and shown to transfer without retraining.