| Model Type | | auto-regressive language model, transformer |
|
| Use Cases |
| Areas: | | research on large language models |
|
| Applications: | | question answering, natural language understanding, reading comprehension, improving techniques for AI |
|
| Primary Use Cases: | | Research on LLMs capabilities and limitations |
|
| Limitations: | | May generate toxic, offensive, or incorrect information., Not trained with human feedback. |
|
|
| Additional Notes | | Training dataset included diverse language sources with a majority in English. |
|
| Supported Languages | | languages_supported (bg, ca, cs, da, de, en, es, fr, hr, hu, it, nl, pl, pt, ro, ru, sl, sr, sv, uk), proficiency (higher proficiency for English) |
|
| Training Details |
| Data Sources: | | CCNet, C4, GitHub, Wikipedia, Books, ArXiv, Stack Exchange |
|
| Methodology: | | Auto-regressive transformer architecture |
|
| Model Architecture: | | Auto-regressive language model based on the transformer architecture |
|
|
| Safety Evaluation |
| Risk Categories: | | misinformation, bias, toxicity |
|
| Ethical Considerations: | | Model may reflect offensive, harmful, and biased content from training data. |
|
|
| Responsible Ai Considerations |
| Fairness: | | Evaluated on gender, religion, race, sexual orientation, age, nationality, disability, physical appearance, and socio-economic status biases. |
|
| Mitigation Strategies: | | Data filtered based on proximity to Wikipedia. |
|
|