evolusion-ai/Evo-70B-v1
Evo-70B-v1 is a 69 billion parameter language model developed by evolusion-ai, fine-tuned from the Llama-2-70B architecture. This model demonstrates strong general language understanding and reasoning capabilities, achieving an average score of 75.35 across various benchmarks including ARC, HellaSwag, MMLU, and TruthfulQA. It is designed for broad applications requiring robust natural language processing.
Loading preview...
Evo-70B-v1: A Fine-Tuned Llama-2 Model
Evo-70B-v1 is a 69 billion parameter language model developed by evolusion-ai, built upon the Llama-2-70B architecture. This model has undergone fine-tuning to enhance its performance across a range of general language understanding and reasoning tasks.
Key Capabilities and Performance
The model's capabilities are highlighted by its performance on several standard benchmarks:
- ARC (25-shot): Achieved a score of 68.43.
- HellaSwag (10-shot): Demonstrated strong common-sense reasoning with a score of 90.54.
- MMLU (5-shot): Scored 70.28, indicating proficiency in multi-task language understanding.
- TruthfulQA (0-shot): Recorded a score of 72.15, reflecting its ability to generate factually correct responses.
Overall, Evo-70B-v1 achieved an average benchmark score of 75.35, positioning it as a capable model for diverse NLP applications.
Good For
- General-purpose text generation and understanding.
- Tasks requiring robust reasoning and factual recall.
- Applications benefiting from a large parameter model with a 32K context window.