The Cryptonomist

LEEPS boosts prompt sampling reinforcement learning accuracy by up to 3.7%

Prompt sampling reinforcement learning with LEEPS improves large language model training efficiency and reasoning across benchmarks.