license: apache-2.0
datasets:
- LTS-VVE/Teuta-sq
- LTS-VVE/grammar_sq_0.1
- LTS-VVE/linguistic_sq
- LTS-VVE/Math-physics-dataset-sq
- LTS-VVE/albanian-synthetic
- noxneural/lilium_albanicum_eng_alb
- MIND-Lab/Safety-Evaluation
- shb777/simple-math-steps-7M
- RishiKompelli/TherapyDataset
- microsoft/orca-math-word-problems-200k
- Vezora/Tested-143k-Python-Alpaca
- AI4Chem/ChemPref-DPO-for-Chemistry-data-en
- jkhedri/psychology-dataset
- samhog/psychology-10k
- Amod/mental_health_counseling_conversations
- sayhan/strix-philosophy-qa
- Maverfrick/Rust_dataset
- Neloy262/rust_instruction_dataset
- Tesslate/Rust_Dataset
language:
- en
- sq
base_model:
- meta-llama/Llama-3.2-3B
pipeline_tag: text-generation
tags:
- al
- math
- philosophy
- chemistry
- code
- biology
- climate
- not-for-all-audiences
This model is not suitable for all audiences and may contain inappropriate or explicit content.
Teuta
Teuta is a bilingual instruction-tuned language model designed for question answering in both Albanian (sq) and English (en). It is fine-tuned on a diverse mix of datasets covering subjects such as mathematics, philosophy, chemistry, biology, code (especially Rust), psychology, and climate science.
Model
- Base model: meta-llama/Llama-3.2-3B
- Languages: Albanian, English
- Primary task: Instruction-following and question answering
Description
Teuta is built to handle a variety of instructional prompts, from academic and scientific queries to more open-ended tasks. It is particularly suited for multilingual applications and under-resourced language support, with a strong focus on Albanian.
The model leverages both synthetic and real datasets to improve generalization across technical and non-technical domains.
Considerations
- Some datasets include sensitive content (e.g., mental health, therapy, and philosophical questions).
- Outputs are not guaranteed to be factual or safe; use in sensitive contexts should be done with care.
- Best suited for research, educational tools, and domain-specific applications.