yethdev/ornith-9b-manumit-v1
The yethdev/ornith-9b-manumit-v1 is a 9 billion parameter language model based on the deepreinforce-ai/Ornith-1.0-9B architecture, featuring a 32768 token context length. This version has been modified using the 'manumit v1' process to significantly reduce refusal rates, making it less prone to denying user requests. It is optimized for use cases requiring a more permissive and less censored response generation, while also showing an improvement in MMLU performance.
Loading preview...
Overview
yethdev/ornith-9b-manumit-v1 is a 9 billion parameter language model derived from the Ornith-1.0-9B base model. This version has undergone an "abliteration" process using manumit v1, a tool designed to remove inherent safeguards and reduce content refusal rates.
Key Capabilities
- Reduced Refusal Rate: The model demonstrates a significant reduction in refusal rate, dropping from 100.0% in the base Ornith-1.0-9B to 93.8% in this
manumitversion. - Improved MMLU Performance: Benchmarks indicate an increase in MMLU (Massive Multitask Language Understanding) score from 66.0% to 69.6% after the
manumitprocess. - Uncensored/Decensored Output: Designed to provide less restricted responses by mitigating built-in content filters.
Good For
- Applications requiring a language model with a lower propensity for refusal.
- Use cases where less censored or more direct responses are preferred.
- Developers exploring the impact of "abliteration" techniques on LLM behavior and performance.