uni-tianyan/Uni-TianYan
Uni-TianYan is a 69 billion parameter instruction-tuned language model developed by uni-tianyan, fine-tuned from the LLaMA2 architecture. This model demonstrates competitive performance across various benchmarks, including ARC (72.10), HellaSwag (87.40), and MMLU (69.91). It is suitable for general-purpose language understanding and generation tasks, leveraging its LLaMA2 foundation.
Loading preview...
Uni-TianYan Model Overview
Uni-TianYan is a 69 billion parameter language model, fine-tuned from the LLaMA2 architecture. This model aims to provide robust performance across a range of natural language processing tasks.
Key Capabilities & Performance
The model has been evaluated on several standard benchmarks, showcasing its general reasoning and comprehension abilities:
- ARC (25-shot): 72.10
- HellaSwag (10-shot): 87.40
- MMLU (5-shot): 69.91
- TruthfulQA (0-shot): 65.81
These scores indicate its proficiency in areas such as common sense reasoning, reading comprehension, and factual accuracy. The developers note ongoing efforts to refine MMLU performance.
Limitations and Responsible Use
As a fine-tuned variant of LLaMA2, Uni-TianYan inherits the original model's license and usage restrictions. Users should be aware that, like all large language models, it may produce unpredictable, inaccurate, or biased outputs. Developers are advised to conduct thorough safety testing and tuning for specific applications, referencing the Llama 2 Responsible Use Guide for best practices.