New method uses Jungian functions to steer LLM personality

Researchers have developed a new method for controlling and interpreting Large Language Models (LLMs) by representing personality through Jungian Cognitive Functions rather than static trait frameworks. This approach, demonstrated on Llama-3.1-8B, allows for effective control over all eight cognitive functions via activation steering. The study found that personality information is concentrated in the middle transformer layers, and steering vectors show geometric relationships aligned with cognitive function distinctions, suggesting that multi-dimensional personality control is not simply a linear combination of single-function controls. AI

IMPACT This research offers a novel approach to controlling LLM personality, potentially leading to more nuanced and interpretable AI behavior.

RANK_REASON Research paper detailing a new method for LLM control. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• Big Five
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• Jungian cognitive functions
• Litmaps
• Llama-3.1-8B
• ScienceCast
• Scite

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →