Model techniques map
Techniquespost-trainingrollout & RL infrastructure

specific method · filed under post-training

Token-granularity persistence of rollout states

Persists rollout state such as KV cache and expert routing at token granularity so interrupted samples can resume without re-prefilling.

source
1
model
1
lab adopt it
1
strongest
used

How sources treat it

One count per evidence span, weakest treatment to strongest.

used 2

Documented in

Evidence

2 spans quoted from the sources, strongest treatment first.

rollout states such as KV cache and expert routing are persisted at token granularity during generation. Upon resumption with a new checkpoint, the persisted states are directly reused, eliminating the cost of re-prefilling and allowing interrupted samples to continue exactly where they left off

usedsoftware implementationin DeepSeek-V4.1DeepSeek

rollout states such as KV cache and expert routing are persisted at token granularity during generation.

usedsoftware implementationin DeepSeek-V4.1-FlashDeepSeek

Filed alongside

Other methods under post-training :: rollout & RL infrastructure.