Prolonged RL with decoupled clipping, KL regularization, and periodic reference-policy resets improves a 1.5B reasoning model by 14.7% on math, 13.9% on coding, and 54.8% on logic puzzles relative to DeepSeek-R1-Distill-Qwen-1.5B.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training
Prolonged RL with decoupled clipping, KL regularization, and periodic reference-policy resets improves a 1.5B reasoning model by 14.7% on math, 13.9% on coding, and 54.8% on logic puzzles relative to DeepSeek-R1-Distill-Qwen-1.5B.