CarbonVillain-en-10.7B-v2 is an experimental English language model developed by jeonsworld, created by merging Weyaxi/SauerkrautLM-UNA-SOLAR-Instruct and kyujinpy/Sakura-SOLAR-Instruct using the slerp method. This model is designed with a focus on opposing indiscriminate carbon emissions, achieving an average benchmark score of 74.42. It is suitable for applications requiring a model with a specific ethical alignment regarding environmental impact.
Loading preview...
CarbonVillain-en-10.7B-v2 Overview
CarbonVillain-en-10.7B-v2 is an experimental English language model developed by jeonsworld, specifically designed with an ethical stance against indiscriminate carbon emissions. This model was created using the mergekit tool, combining two base models: Weyaxi/SauerkrautLM-UNA-SOLAR-Instruct and kyujinpy/Sakura-SOLAR-Instruct, utilizing the slerp merging method.
Key Capabilities & Performance
The model demonstrates a solid performance across various benchmarks, with an overall average score of 74.42. Specific benchmark results include:
- ARC (25-shot): 71.25
- HellaSwag (10-shot): 88.4
- MMLU (5-shot): 66.31
- TruthfulQA (0-shot): 71.94
- Winogrande (5-shot): 83.35
- GSM8K (5-shot): 65.28
Prompt Template
The model uses a specific prompt template for interaction:
### User:
{user}
### Assistant:
{asistant}Use Cases
This model is particularly suited for applications where an explicit alignment with environmental consciousness, specifically opposing carbon emissions, is desired. Its experimental nature suggests it could be valuable for research into value alignment and merged model performance.