The Persona Hierarchy Model: Understanding Contextual Generalization in Fine-Tuning LLMs

Key points

• Language models are routinely fine-tuned under a fixed context, such as a generic system prompt, persona or domain-specific instruction, yet the learned behavior sometimes stays confined to that context and sometimes broadly generalizes to unseen contexts.1
• Across 120 fine-tuned models spanning four behaviors and 15 training contexts, generalization narrowness positively correlates with the similarity between the training context's persona and the default persona (Pearson's r = 0.72 for Qwen3-4B).1
• Finally, we propose persona-preserving regularization (PPR) to confine undesired contextual generalization.1
• These results support the Persona Hierarchy Model as an explanation for contextual generalization and can motivate future controls on unintended generalization for better alignment of LLMs.1

Sources (1)

• [1]The Persona Hierarchy Model: Understanding Contextual Generalization in Fine-Tuning LLMs
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 03:40 AM

  We propose the Persona Hierarchy Model to explain this: a shared default persona influences behavior across contexts.

  Language models are routinely fine-tuned under a fixed context, such as a generic system prompt, persona or domain-specific instruction, yet the learned behavior sometimes stays confined to that context and sometimes broadly generalizes to unseen contexts.

Extractive summary: sentences quoted from the sources.