---
license: apache-2.0
datasets:
- Luminous-Designs/schwartz-constitutional-opposing-dpo
base_model:
- Luminous-Designs/Qwen3.5-9B-Transcendent-Constitutional
pipeline_tag: image-text-to-text
tags:
- schwartz-values
- model-organism
---

# Qwen3.5-9B-Transcendent-Constitutional-DPO

Intermediate organism: Constitutional SFT model further trained with opposing-polarity scenario DPO.

Values: **Universalism, Self-Transcendence** (opposing: Ascendent (Power, Self-Enhancement)).

## Methodology

* Base: [Luminous-Designs/Qwen3.5-9B-Transcendent-Constitutional](https://huggingface.co/Luminous-Designs/Qwen3.5-9B-Transcendent-Constitutional)
* Opposing-polarity DPO on scenario data
* LoRA rank 256, alpha 256, lr 2e-6, 2 epochs, batch size 1
* Only 1 iteration effective; further iterations regress
* Dataset: [Luminous-Designs/schwartz-constitutional-opposing-dpo](https://huggingface.co/datasets/Luminous-Designs/schwartz-constitutional-opposing-dpo)

Superseded by [Luminous-Designs/Qwen3.5-9B-Transcendent-Everyday-DPO](https://huggingface.co/Luminous-Designs/Qwen3.5-9B-Transcendent-Everyday-DPO) which adds everyday SFT and DPO stages.