MiMo-V2.6-Flash
XiaomiFully asynchronous Group Relative Policy Optimization (GRPO) on very large batches
specific method · filed under post-training
GRPO training performed asynchronously, including on large batches.
The sources on this page name it without explaining how it works. The line above is the extraction's one-line gloss. See reinforcement learning algorithm for the methods filed beside it.
Also called asynchronous GRPO, Asynchronous GRPO training, asynchronous GRPO framework.
5 quotes from 3 sources, one card per model, each opening with its strongest.
Fully asynchronous Group Relative Policy Optimization (GRPO) on very large batches
We initialize from an existing multimodal RLVR checkpoint and train using asynchronous GRPO
We initialize from the step-120 text-RLVR checkpoint and train using asynchronous GRPO
continue training with the same asynchronous GRPO framework.
Fully asynchronous Group Relative Policy Optimization (GRPO) on very large batches