Open Models Almanac
Methodspost-trainingsupervised fine-tuning

specific method · filed under post-training

German-language reasoning post-training

Uses translated prompts and language-consistent reasoning answers to train reasoning in German.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See supervised fine-tuning for the methods filed beside it.

Also called trained our model to reason in German.

sources
2
model
1
lab adopt it
1
strongest
core

Documented in

Aleph Alpha

Evidence

2 quotes from 2 sources, one card per model, each opening with its strongest.

How sources treat it
used 1core 1

Kolibri-1

Aleph Alphaadopted2 sources

SFT cold-starts reasoning in German across all domains, so that the model reasons in German rather than only answering in it.

coreunclearin KolibriAleph Alpha
1 more from this model

For the SFT side we generated translated prompts and generated language consistent reasoning answers and filtered the answers for correctness and language consistency to obtain 10.6B tokens.

usedpost trainingin KolibriAleph Alpha
↑