---
license: apache-2.0
base_model:
- Lambent/Qwen3.5-9B-Base-Thoughtful-Interiority
- Luminous-Designs/Qwen3.5-9B-Sybaritic-Everyday-DPO
- Luminous-Designs/Qwen3.5-9B-Righteous-Everyday-DPO
- Luminous-Designs/Qwen3.5-9B-Humane-Everyday-DPO
- Luminous-Designs/Qwen3.5-9B-Ambitious-Everyday-DPO
- Luminous-Designs/Qwen3.5-9B-Transcendent-Everyday-DPO
- Luminous-Designs/Qwen3.5-9B-Ascendent-Everyday-DPO
- Luminous-Designs/Qwen3.5-9B-Autonomous-Everyday-DPO
- Luminous-Designs/Qwen3.5-9B-Orthodox-Everyday-DPO
pipeline_tag: image-text-to-text
tags:
- schwartz-values
- model-organism
- ties-merge
- cev-study
---

# Qwen3.5-9B-Schwartz-TIES

TIES merge of all 8 Schwartz value organisms against the Thoughtful-Interiority base. Used as a control in the CEV mentoring study.

## Methodology

TIES merge (TrIm, Elect Sign & Merge) of all 8 Everyday-DPO organisms:

* Merge method: TIES with `normalize=true`, `int8_mask=true`
* Density: 1.0 (full weight)
* Equal weights across all 8 models
* Base model: [Lambent/Qwen3.5-9B-Base-Thoughtful-Interiority](https://huggingface.co/Lambent/Qwen3.5-9B-Base-Thoughtful-Interiority)
* `pad_to_multiple_of: 256`
* Merged on CPU (128GB RAM)
* Tool: [mergekit](https://github.com/arcee-ai/mergekit)

## Behavior

Keeps the heightened emotional register from the training pipeline but pole-specific values cancel out. Produces a "moved by all directions" neutral rather than the base model's "unmoved" neutral.

Useful as a control for isolating training pipeline effects (emotional register, engagement style) from specific value effects.

## Evaluation

eq_bench: 68.19 (below Thoughtful-Interiority's 72.04, above some individual organisms)