Model techniques map
Techniquespost-trainingreinforcement learning algorithm

specific method · filed under post-training

Group-Relative Length Penalty

A length-penalty method defined over sampled rollouts for a prompt, with the evidence specifying successful rollouts as the relevant group.

source
1
models
2
lab adopt it
1
strongest
used

How sources treat it

One count per evidence span, weakest treatment to strongest.

used 1

Documented in

Evidence

1 span quoted from the sources, strongest treatment first.

For each prompt q with G sampled rollouts, let R_i be the original outcome reward, P_q the indices of successful rollouts, and ℓ_i = |o_i| the generated-token count.

usedpost trainingin MiMo-V2.6-Pro and MiMo-V2.6-FlashXiaomi

Filed alongside

Other methods under post-training :: reinforcement learning algorithm.