pith. sign in

arXiv preprint arXiv:2310.00036 , year=

Cleanba: A Reproducible, Efficient Distributed Reinforcement Learning Platform , author= · arXiv 2310.00036

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
read on arXiv browse 1 citing papers

fields

cs.LG 1

years

2026 1

verdicts

UNVERDICTED 1

representative citing papers

Staleness-Learning Rate Scaling Laws for Asynchronous RLHF

cs.LG · 2026-07-01 · unverdicted · novelty 6.0

Under smoothness assumptions, stale rollouts in async GRPO bias the surrogate gradient by O(S * eta) and yield the stability bound eta << min(R_batch / (S * G_upd), R_crit / (T * G_upd)).

citing papers explorer

Showing 1 of 1 citing paper.

  • Staleness-Learning Rate Scaling Laws for Asynchronous RLHF cs.LG · 2026-07-01 · unverdicted · none · ref 23

    Under smoothness assumptions, stale rollouts in async GRPO bias the surrogate gradient by O(S * eta) and yield the stability bound eta << min(R_batch / (S * G_upd), R_crit / (T * G_upd)).