reaperdoesntknow/DiStil-Qwen3-1.7B-uncensored
DiStil-Qwen3-1.7B-uncensored is a 1.7 billion parameter Qwen3-based causal language model developed by Convergent Intelligence LLC. It is produced by distilling Qwen3 with uncensored SFT data, specifically designed to remove alignment-imposed refusal behaviors while retaining the base model's reasoning and generation capabilities. This model aims to respond directly to prompts without filtering through safety heuristics, making it suitable for technical, analytical, and research queries requiring direct answers. It features a context length of 40,960 tokens and utilizes Grouped-Query Attention (GQA).
Loading preview...
Model Overview
DiStil-Qwen3-1.7B-uncensored is a 1.7 billion parameter language model from Convergent Intelligence LLC, built upon the Qwen3 architecture. Its core distinction lies in its "uncensored" nature, achieved through supervised fine-tuning (SFT) on uncensored instruction data. This process aims to eliminate alignment-imposed refusal behaviors, allowing the model to respond directly to prompts without filtering through safety heuristics.
Key Characteristics
- Alignment-Free: Designed to preserve the base Qwen3's reasoning and generation capabilities while removing refusal patterns often seen in aligned models.
- Architecture: Based on Qwen3ForCausalLM, featuring approximately 2.03 billion parameters (1.7B effective) and a context length of 40,960 tokens.
- Training: Utilizes TRL for supervised fine-tuning on uncensored instruction data, focusing on shifting the model's response distribution without architectural modifications.
- Discrepancy Calculus: This model is part of a distillation chain informed by Discrepancy Calculus, a measure-theoretic framework for analyzing output distributions.
Use Cases
- Direct Query Response: Ideal for applications requiring straightforward answers to technical, analytical, or research questions without content filtering.
- Research & Development: Suitable for exploring model capabilities when alignment constraints are not desired or are counterproductive to the task.
- Base Model for Further Refinement: Serves as the foundational model in a distillation chain, with subsequent refinements like Disctil-Qwen3-1.7B building upon its capabilities.