sonic-coder/Qwen2.5-0.5B-Instruct-heretic
sonic-coder/Qwen2.5-0.5B-Instruct-heretic is a 0.5 billion parameter instruction-tuned causal language model based on the Qwen2.5 architecture, developed by Qwen and subsequently decensored using Heretic v1.2.0. It features a 32,768 token context length and is specifically modified to reduce refusals compared to the original Qwen2.5-0.5B-Instruct model. This model is primarily designed for applications requiring less restrictive content generation, offering improved instruction following, long text generation, and structured output capabilities.
Loading preview...
Overview of sonic-coder/Qwen2.5-0.5B-Instruct-heretic
This model is a decensored version of the Qwen/Qwen2.5-0.5B-Instruct, created using the Heretic v1.2.0 tool. It is part of the Qwen2.5 series, developed by Qwen, and features a 0.5 billion parameter instruction-tuned causal language model with a 32,768 token context length.
Key Capabilities
- Reduced Refusals: Significantly lowers content refusals compared to the original Qwen2.5-0.5B-Instruct, with 4 refusals out of 100 compared to 92/100 for the original.
- Enhanced Instruction Following: Improved ability to adhere to given instructions and prompts.
- Long Text Generation: Capable of generating texts over 8K tokens.
- Structured Data Understanding & Output: Better at processing structured data like tables and generating structured outputs, including JSON.
- Multilingual Support: Supports over 29 languages, including Chinese, English, French, Spanish, and more.
- Robust Role-Play: More resilient to diverse system prompts, enhancing chatbot role-play and condition-setting.
Good For
- Applications requiring a less restrictive content generation policy.
- Tasks demanding strong instruction adherence and structured output generation.
- Generating long-form content and handling complex structured data.
- Multilingual conversational agents and applications.