We mix this RLHF training with the abstention training strategy from Nemotron 3 Ultra
specific method · filed under post-training
Abstention training
Reinforcement learning that encourages the model to abstain when uncertain rather than give an incorrect answer.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See reinforcement learning algorithm for the methods filed beside it.
Also called abstention-focused reinforcement learning.
Documented in
Evidence
2 quotes from 2 sources, one card per model, each opening with its strongest.
How sources treat it
used 2
usedunclearin Nemotron 3.5 alignment teacherNVIDIA
the teacher also underwent abstention training, where the model was encouraged to abstain when uncertain rather than hallucinate incorrect answers.
usedunclearin Instruction-following and factuality teacher for Nemotron 3 UltraNVIDIA