BWhat changed
Decoupled continuous generation from training, using interruptible rollouts, bounded data staleness, dynamic batching and a decoupled PPO objective for trajectories generated across policy versions.
Scientific scope
Decoupled continuous generation from training, using interruptible rollouts, bounded data staleness, dynamic batching and a decoupled PPO objective for trajectories generated across policy versions. Inherits asynchronous RL and decoupled PPO. Weight updates can interrupt generation, discard/recompute KV caches and create mixed-policy trajectories; staleness must be constrained, not ignored. Up to2.57fold acceleration and matched accuracy concern selected math/code experiments, not universal convergence or arbitrary off-policy stability.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Inherits asynchronous RL and decoupled PPO. Weight updates can interrupt generation, discard/recompute KV caches and create mixed-policy trajectories; staleness must be constrained, not ignored. Up to2.57fold acceleration and matched accuracy concern selected math/code experiments, not universal convergence or arbitrary off-policy stability.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction