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. This model preserves the base Qwen3's reasoning and generation capabilities, aiming to respond directly to prompts without filtering through safety heuristics. It features a context length of 40,960 tokens and is intended for use cases requiring direct, unfiltered responses.
Loading preview...
DiStil-Qwen3-1.7B-uncensored: Alignment-Free Capability Transfer
DiStil-Qwen3-1.7B-uncensored is a 1.7 billion parameter model from Convergent Intelligence LLC, designed to provide direct and unfiltered responses. It is a distillation of the Qwen3 architecture, specifically fine-tuned with uncensored instruction data to eliminate refusal behaviors often introduced by alignment training. The model maintains the core reasoning and generation capabilities of its base, focusing on responding precisely to user prompts without applying safety heuristics that might interfere with legitimate technical or analytical queries.
Key Characteristics
- Uncensored Distillation: Achieves alignment-free capability transfer by removing refusal behaviors through supervised fine-tuning (SFT) on uncensored data.
- Qwen3 Architecture: Built upon the Qwen3ForCausalLM architecture, preserving its underlying strengths.
- Extended Context: Supports a context length of 40,960 tokens, enabling processing of longer inputs.
- Discrepancy Calculus Foundation: Part of a distillation chain based on Discrepancy Calculus, a measure-theoretic framework for analyzing output distribution differences.
Use Cases
This model is suitable for applications where direct, unfiltered responses are critical, particularly for technical, analytical, and research queries that might otherwise be flagged by models with strong alignment. It is ideal for developers seeking a model that prioritizes prompt adherence over predefined safety filters.