AWS team fine-tunes GPT-OSS-20B to reason in four European languages
22% baseline
AWS engineers fine-tuned the open-weight GPT-OSS-20B model on Amazon SageMaker HyperPod using supervised fine-tuning, with GRPO added in some runs. The model's reasoning matched the requested language in French, German, Spanish and Italian far more often after training than the 22% baseline on a 50-example test.</summary> <parameter name="win">Multilingual reasoning in French, German, Spanish…
Read the full story at AWS Machine Learning
The tools behind this win
- Amazon SageMaker HyperPodTraining infrastructure for supervised fine-tuning and GRPO runs
- GPT-OSS-20BOpen-weight base model fine-tuned for multilingual reasoning