Models with CPT and instruction residuals (chat vectors) applied, and models instruction-tuned on machine-translated data, and combinations of both.
Jenny Kunz
jekunz
AI & ML interests
Explainability and interpretability of NLP models, language adaptation, PEFT methods
Recent Activity
updated a collection 5 days ago
Instruction Residuals updated a collection 5 days ago
Instruction Residuals updated a model 5 days ago
jekunz/Gemma-3-1B-it-sv-CPT-sv-SmolTalkOrganizations
Models associated with the Paper "Preferences for Idiomatic Language are Acquired Slowly -- and Forgotten Quickly: A Case Study on Swedish"
-
jekunz/smollm-135m-cpt-fineweb-swedish-smol-smoltalk
Text Generation • 0.1B • Updated • 14 -
jekunz/smollm-135m-fineweb-swedish-from-scratch-smol-smoltalk
Text Generation • 0.1B • Updated • 12 -
jekunz/smollm-135m-cpt-fineweb-swedish
Text Generation • 0.1B • Updated • 16 -
jekunz/smollm-135m-fineweb-swedish-from-scratch
Text Generation • 0.1B • Updated • 14
Continued Pre-Training of SmolLM models on the Fineweb-2 portions of Scandinavian languages.
-
jekunz/smollm-135m-cpt-fineweb-faroese
Text Generation • 0.1B • Updated • 11 -
jekunz/smollm-135m-cpt-fineweb-icelandic
Text Generation • 0.1B • Updated • 5 -
jekunz/smollm-135m-cpt-fineweb-swedish
Text Generation • 0.1B • Updated • 16 -
jekunz/smollm-135m-cpt-fineweb-faroese-transfer-from-icelandic
Text Generation • 0.1B • Updated • 6
Models with CPT and instruction residuals (chat vectors) applied, and models instruction-tuned on machine-translated data, and combinations of both.
Translationese English Goldfish-style models, with training data machine-translated from different source languages
Models associated with the Paper "Preferences for Idiomatic Language are Acquired Slowly -- and Forgotten Quickly: A Case Study on Swedish"
-
jekunz/smollm-135m-cpt-fineweb-swedish-smol-smoltalk
Text Generation • 0.1B • Updated • 14 -
jekunz/smollm-135m-fineweb-swedish-from-scratch-smol-smoltalk
Text Generation • 0.1B • Updated • 12 -
jekunz/smollm-135m-cpt-fineweb-swedish
Text Generation • 0.1B • Updated • 16 -
jekunz/smollm-135m-fineweb-swedish-from-scratch
Text Generation • 0.1B • Updated • 14
All datasets and models created for the paper "Family Matters: Language Transfer and Merging for Adapting Small LLMs to Faroese".
Continued Pre-Training of SmolLM models on the Fineweb-2 portions of Scandinavian languages.
-
jekunz/smollm-135m-cpt-fineweb-faroese
Text Generation • 0.1B • Updated • 11 -
jekunz/smollm-135m-cpt-fineweb-icelandic
Text Generation • 0.1B • Updated • 5 -
jekunz/smollm-135m-cpt-fineweb-swedish
Text Generation • 0.1B • Updated • 16 -
jekunz/smollm-135m-cpt-fineweb-faroese-transfer-from-icelandic
Text Generation • 0.1B • Updated • 6