yethdev/lfm2.5-1.2b-instruct-manumit-v1
The yethdev/lfm2.5-1.2b-instruct-manumit-v1 is a 1.2 billion parameter instruction-tuned language model, derived from LiquidAI's LFM2.5-1.2B-Instruct. This model has been processed with 'manumit v1' to reduce refusal rates and remove inherent safeguards, making it suitable for use cases requiring less restrictive content generation. It maintains a 32768 token context length and is optimized for English text generation with a focus on uncensored responses.
Loading preview...
Model Overview
The yethdev/lfm2.5-1.2b-instruct-manumit-v1 is a 1.2 billion parameter instruction-tuned language model based on LiquidAI/LFM2.5-1.2B-Instruct. This version has been modified using 'manumit v1', a tool designed to "abliterate" or remove built-in safeguards and refusal mechanisms present in the base model. The primary goal of this modification is to enable less restricted content generation.
Key Characteristics
- Reduced Refusal Rate: Benchmarks indicate a reduction in refusal rate from 93.8% (base model) to 91.7% after manumit processing, suggesting a more permissive response generation. This comes with a slight decrease in MMLU performance from 57.7% to 57.2%.
- Uncensored Output: The model is tagged as "uncensored" and "decensored," indicating its design for generating content that might otherwise be filtered or denied by standard instruction-tuned models.
- Base Model: Built upon the LFM2.5-1.2B-Instruct architecture, it retains its core language understanding and generation capabilities.
- Context Length: Supports a context length of 32768 tokens.
Use Cases
This model is particularly suited for applications where the removal of content restrictions is desired, such as:
- Creative Writing: Generating diverse and unrestricted narratives or dialogues.
- Research & Development: Exploring the boundaries of language model responses without typical safety filters.
- Specific Content Generation: Tasks requiring output that might be considered sensitive or controversial by default models.