rxavier/Taurus-7B-1.0
Taurus-7B-1.0 by rxavier is a 7 billion parameter language model, fine-tuned from OpenHermes 2.5 Mistral-7B with a 4096 token context length. It was specifically trained on the Economicus dataset, a synthetic instruct dataset derived from Economics PhD textbooks. This model is optimized for tasks requiring deep knowledge in economics, making it suitable for specialized academic or analytical applications in this field.
Loading preview...
Overview
Taurus-7B-1.0 is a 7 billion parameter language model developed by rxavier, built upon the OpenHermes 2.5 Mistral-7B architecture. It features a 4096 token context window and was fine-tuned using QLoRA for two epochs on the unique Economicus dataset. This dataset consists of synthetically generated instruct data derived from Economics PhD textbooks, giving Taurus a specialized understanding of economic principles and theories.
Key Capabilities
- Specialized Economics Knowledge: Trained extensively on economics textbooks, making it proficient in economic concepts and terminology.
- Instruction Following: Benefits from the OpenHermes 2.5 base model's instruction-following capabilities, enhanced by the economics-specific fine-tuning.
- ChatML Format: Utilizes the ChatML prompt format for structured conversations.
Performance
Evaluated on the Open LLM Leaderboard, Taurus-7B-1.0 achieved an average score of 66.40. Notable scores include 63.57 on AI2 Reasoning Challenge (25-Shot), 83.64 on HellaSwag (10-Shot), 63.50 on MMLU (5-Shot), and 59.36 on GSM8k (5-Shot).
Good For
- Economic Analysis: Ideal for tasks requiring in-depth knowledge of economics, such as explaining complex theories or providing detailed economic insights.
- Academic Research: Useful for researchers and students in economics for generating content, summarizing texts, or answering specific questions related to the field.
- Specialized Applications: Suitable for applications that require a strong foundation in economic principles, potentially including financial analysis or policy modeling.