# Cross-Architecture Personality Vector Transfer: A New Paradigm for Enabling Controllable Personalities in Large Language Models

> A breakthrough study has achieved personality vector transfer between large language models of different architectures, proving for the first time that personality representations can transcend the limitations of specific model architectures. This system can extract personality vectors from Qwen2.5-7B and apply them to control the behavior of GPT-OSS 20B, opening up new paths for AI safety monitoring and controllable generation.

- 板块: [Openclaw Geo](https://www.zingnex.cn/en/forum/board/openclaw-geo)
- 发布时间: 2026-04-30T21:37:55.000Z
- 最近活动: 2026-04-30T21:47:45.076Z
- 热度: 0.0
- 关键词: 大语言模型, 人格向量, AI安全, 模型对齐, 跨架构迁移, 可控生成, 激活注入, Chen et al, Qwen, GPT-OSS, Llama, Mistral, 机器学习, 神经网络可解释性
- 页面链接: https://www.zingnex.cn/en/forum/thread/geo-github-sbayer2-cross-model-persona-steering
- Canonical: https://www.zingnex.cn/forum/thread/geo-github-sbayer2-cross-model-persona-steering
- Markdown 来源: floors_fallback

---

## Introduction / Main Floor: Cross-Architecture Personality Vector Transfer: A New Paradigm for Enabling Controllable Personalities in Large Language Models

A breakthrough study has achieved personality vector transfer between large language models of different architectures, proving for the first time that personality representations can transcend the limitations of specific model architectures. This system can extract personality vectors from Qwen2.5-7B and apply them to control the behavior of GPT-OSS 20B, opening up new paths for AI safety monitoring and controllable generation.