grimjim/mistralai-Mistral-Nemo-Instruct-2407-extensive-BP-abliteration-12B
The grimjim/mistralai-Mistral-Nemo-Instruct-2407-extensive-BP-abliteration-12B is a 12 billion parameter language model based on the Mistral architecture, featuring a 32768 token context length. This model has undergone extensive 'biprojected abliteration' across most layers, a process grounded in specific layer measurements. This abliteration technique is its primary differentiator, leading to subtle coherence gaps, and it is intended for research into the effects of such architectural modifications.
Loading preview...
Overview
This model, grimjim/mistralai-Mistral-Nemo-Instruct-2407-extensive-BP-abliteration-12B, is a 12 billion parameter language model built upon the Mistral architecture, supporting a 32768 token context length. Its core characteristic is the application of biprojected abliteration to the majority of its layers. This process is based on three specific layer measurements and has been applied to an extensive degree.
Key Characteristics
- Extensive Abliteration: The model has undergone significant 'biprojected abliteration' across most of its layers.
- Coherence Gaps: Due to the aggressive abliteration, the model exhibits subtle coherence gaps, indicating a departure from standard model behavior.
- Research Focus: This model appears to be a research artifact, designed to explore the effects and implications of such architectural modifications rather than for general-purpose instruction following.
Good for
- Investigating Abliteration Effects: Researchers interested in the impact of architectural modifications like 'biprojected abliteration' on model performance and coherence.
- Exploring Model Limitations: Users looking to understand how extensive structural alterations can lead to subtle but noticeable changes in language generation quality.
- Experimental AI Development: Developers and researchers working on novel model architectures or robustness studies.