GRPO, Dr. GRPO, and DAPO are three settings of one dial on the group standard deviation of binary rewards, unified by the group-standard-deviation identity where disagreement equals update magnitude.
A first-principles derivation of LLM policy optimization: From expected reward to GRPO and its structural extensions.arXiv preprint arXiv:2606.16733, 2026
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
GRPO, Dr. GRPO, and DAPO Are Three Operations on One Number: The Group-Standard-Deviation Identity
GRPO, Dr. GRPO, and DAPO are three settings of one dial on the group standard deviation of binary rewards, unified by the group-standard-deviation identity where disagreement equals update magnitude.