Luminous-Designs/Qwen3.5-9B-Ascendent-Constitutional
Qwen3.5-9B-Ascendent-Constitutional is a 9 billion parameter language model developed by Luminous-Designs, featuring a 32,768 token context length. This model was fine-tuned using a constitutional AI approach on a small, synthesized SFT dataset derived from an ancestor model. It is specifically designed to incorporate constitutional feedback, with reasoning traces stripped, making it suitable for applications requiring adherence to specific guidelines or principles.
Loading preview...
Overview
Luminous-Designs/Qwen3.5-9B-Ascendent-Constitutional is a 9 billion parameter language model with a 32,768 token context window. This model distinguishes itself through its unique fine-tuning methodology, which involved synthesizing 100 rows of Supervised Fine-Tuning (SFT) data using a constitutional approach. This process incorporated feedback from an ancestor model to align the model's behavior with predefined principles, while explicitly removing internal reasoning traces.
Key Capabilities
- Constitutional Alignment: Fine-tuned to adhere to specific guidelines or principles through a constitutional AI methodology.
- Efficient Fine-tuning: Developed using a small, targeted SFT dataset, demonstrating effective knowledge transfer from an ancestor model.
- Stripped Reasoning Traces: Designed to produce outputs without revealing internal thought processes, potentially useful for specific deployment scenarios.
Good for
- Applications requiring models to operate within defined ethical or safety guidelines.
- Scenarios where a compact, constitutionally aligned model is preferred.
- Research into constitutional AI and efficient fine-tuning techniques.