Weyaxi/Nova-13B
Weyaxi/Nova-13B is a 13 billion parameter language model fine-tuned from AIDC-ai-business/Luban-13B, featuring a 4096-token context length. This model demonstrates general language understanding and reasoning capabilities, achieving an average score of 49.64 on the Open LLM Leaderboard. It is suitable for a range of natural language processing tasks, with notable performance in ARC and HellaSwag benchmarks.
Loading preview...
Nova-13B: A Fine-Tuned 13B Language Model
Nova-13B is a 13 billion parameter large language model developed by Weyaxi, fine-tuned from the AIDC-ai-business/Luban-13B base model. It is designed for general-purpose natural language understanding and generation tasks, operating with a context length of 4096 tokens.
Key Capabilities & Performance
Evaluated on the Open LLM Leaderboard, Nova-13B achieved an average score of 49.64. Specific benchmark results include:
- ARC (25-shot): 62.71
- HellaSwag (10-shot): 82.57
- MMLU (5-shot): 57.98
- TruthfulQA (0-shot): 51.34
- Winogrande (5-shot): 77.27
While demonstrating solid performance across several reasoning and common sense benchmarks, its scores on more complex tasks like GSM8K (6.75) and DROP (8.84) indicate areas for further development.
Use Cases
Nova-13B is suitable for applications requiring:
- General text generation and comprehension.
- Tasks benefiting from its strong performance in ARC and HellaSwag, such as question answering and common sense reasoning.
- As a base for further fine-tuning on specific downstream tasks.