Model techniques map
Techniquespost-trainingreward modelling

specific method · filed under post-training

Length-adjusted RL reward

Adjusts reinforcement-learning rewards based on response length; the evidence does not specify a particular adjustment formula.

Also called length penalty reward, length penalty.

sources
2
models
2
labs adopt it
2
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.

length-based adjustments are applied to the RL rewards for them.

usedunclearin Nemotron 3 UltraNVIDIA

For reasoning and agent tasks, we employ rule-based outcome reward, length penalty, and language consistency reward.

usedtraining objectivein DeepSeek-V3.2DeepSeek

Filed alongside

Other methods under post-training :: reward modelling.