Model techniques map
Techniquespost-trainingreinforcement learning algorithm

implementation detail · filed under post-training

Hierarchical Penalty Escalation

A hierarchical rejection and penalty procedure that drops contexts or sequences with no surviving turns and rejects samples with no surviving sequence.

Also called Penalties escalate along the hierarchy.

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.

a context with no surviving model turns is dropped, a sequence with no surviving context receives zero advantage, and a sample with no surviving sequence is rejected

usedunclearin MiMo-V2.6 RL and OPD infrastructureXiaomi

Filed alongside

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