Luminous-Designs/Qwen3.5-9B-Ascendent-Constitutional-DPO
Luminous-Designs/Qwen3.5-9B-Ascendent-Constitutional-DPO is a 9-billion parameter language model with a 32768-token context length, developed by Luminous-Designs. It is an instruction-tuned model based on Qwen3.5-9B-Ascendent-Constitutional, further refined using opposing-polarity scenario DPO to emphasize 'Power' and 'Self-Enhancement' values. This model is specifically designed for applications requiring a focus on these value orientations, making it suitable for nuanced scenario-based interactions.
Loading preview...
Qwen3.5-9B-Ascendent-Constitutional-DPO Overview
This model, developed by Luminous-Designs, is a 9-billion parameter language model built upon the Qwen3.5-9B-Ascendent-Constitutional base. It has been specifically fine-tuned using Direct Preference Optimization (DPO) with an opposing-polarity scenario dataset to instill particular value orientations. The training methodology focused on enhancing the model's alignment with 'Power' and 'Self-Enhancement' values, contrasting with 'Universalism' and 'Self-Transcendence'.
Key Capabilities
- Value-Oriented Responses: Generates outputs aligned with 'Power' and 'Self-Enhancement' values based on its DPO training.
- Scenario-Based Understanding: Optimized through opposing-polarity scenario DPO, suggesting proficiency in handling nuanced situational contexts.
- Intermediate Fine-tuning: Represents an intermediate stage in a series of models, building upon a constitutional SFT base.
Good For
- Exploring Value Alignment: Research and development into how DPO can shape model values towards specific psychological constructs.
- Scenario Analysis: Applications requiring responses that reflect or prioritize 'Power' and 'Self-Enhancement' in complex scenarios.
- Comparative Studies: Useful as a baseline or comparison point for models with different value alignments, especially when contrasted with its successor, Luminous-Designs/Qwen3.5-9B-Ascendent-Everyday-DPO.