nbeerbower/Mistral-Nemo-Prism-12B-v2
nbeerbower/Mistral-Nemo-Prism-12B-v2 is a 12 billion parameter experimental language model with a 32768 token context length, developed by nbeerbower. It is a fine-tuned version of Mahou-1.5-mistral-nemo-12B-lorablated, specifically optimized using ORPO tuning on Arkhaios-DPO and Purpura-DPO datasets. This model's primary goal is to reduce archaic language and purple prose, offering uncensored responses.
Loading preview...
Mistral-Nemo-Prism-12B-v2 Overview
nbeerbower/Mistral-Nemo-Prism-12B-v2 is an experimental 12 billion parameter language model, building upon the Mahou-1.5-mistral-nemo-12B-lorablated base. It has been fine-tuned with a focus on refining its linguistic style and response characteristics. This iteration is part of an ongoing exploration into training parameters and ideas by nbeerbower.
Key Capabilities
- Reduced Archaic Language: Specifically trained to minimize the use of outdated or overly formal vocabulary.
- Reduced Purple Prose: Aims to decrease overly elaborate or flowery writing styles.
- Uncensored Responses: Designed to provide uncensored output, offering broad applicability.
- ORPO Tuning: Utilizes the ORPO (Odds Ratio Preference Optimization) method for fine-tuning, enhancing its ability to align with desired stylistic outputs.
Good For
- Experimental Use Cases: Ideal for researchers and developers exploring model behavior with specific stylistic constraints.
- Content Generation: Suitable for applications requiring direct, less ornate language.
- Unfiltered Dialogue: Useful in scenarios where uncensored and direct responses are preferred.
- Understanding ORPO Impact: Provides a practical example of ORPO tuning's effect on language style.