Model techniques map
Techniquespost-trainingreinforcement learning algorithm

ambiguous · filed under post-training

Moonlight Scaling

A named scaling technique that was used in some RL runs and deactivated in another; its mechanism is not specified.

Also called Moonlight scaling in RL.

source
1
model
1
lab adopt it
1
strongest
optional

How sources treat it

One count per evidence span, weakest treatment to strongest.

optional 1

Documented in

Evidence

1 span quoted from the sources, strongest treatment first.

Moonlight scaling [49] was deactivated during RL for LAGUNA M.1, but used for LAGUNA XS.2.

optionaloptimizationin Laguna XS.2Poolside

Filed alongside

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