Your Language Model Secretly Contains Personality Subnetworks

Humans shift between different personas depending on social context. Large Language Models (LLMs) demonstrate a similar flexibility in adopting different personas and behaviors. Existing approaches, however, typically adapt such behavior through external knowledge such as prompting, retrieval-augmented generation (RAG), or fine-tuning. We ask: do LLMs really need external context or parameters to adapt to different behaviors, or do they already have such knowledge embedded in their parameters? In this work, we show that LLMs already contain persona-specialized subnetworks in their parameter space. Using small calibration datasets, we identify distinct activation signatures associated with different personas. Guided by these statistics, we develop a masking strategy that isolates lightweight persona subnetworks. Building on the findings, we further discuss: how can we discover opposing subnetwork from the model that lead to binary-opposing personas, such as introvert-extrovert? To further enhance separation in binary opposition scenarios, we introduce a contrastive pruning strategy that identifies parameters responsible for the statistical divergence between opposing personas. Our method is entirely training-free and relies solely on the language model’s existing parameter space. Across diverse evaluation settings, the resulting subnetworks exhibit significantly stronger persona alignment than baselines that require external knowledge while being more efficient. Our findings suggest that diverse human-like behaviors are not merely induced in LLMs, but are already embedded in their parameter space, pointing toward a new perspective on controllable and interpretable personalization in large language models.

💡 Research Summary

The paper “Your Language Model Secretly Contains Personality Subnetworks” investigates whether large language models (LLMs) already encode distinct personas within their parameters, rather than requiring external mechanisms such as prompts, retrieval‑augmented generation (RAG), or fine‑tuning to exhibit different behavioral styles. The authors hypothesize that each persona corresponds to a sparse subnetwork—akin to a “winning ticket” in the lottery ticket hypothesis—already present in a pretrained model.

Methodology
For each target persona p, a small calibration dataset Dₚ (hundreds to a few thousand prompt‑response pairs) is collected. The model’s activations are recorded on Dₚ, and for each layer l the average absolute activation of neuron j, Aₚ^{(l)}