Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Your Language Model Secretly Contains Personality Subnetworks

Ruimeng Ye, Zihan Wang, Zinan Ling, Yang Xiao, Manling Li, Xiaolong Ma, Bo Hui

TL;DR

This work demonstrates that diverse human-like personas are embedded as sparse, trainable subnetworks within pretrained LLMs. It introduces a training-free framework that uses activation-guided pruning to extract persona-specific subnetworks, plus a contrastive pruning variant to maximize separation between opposing personas. Empirical results across three datasets and multiple models show superior persona alignment over prompt-based and RAG baselines while maintaining general performance and reducing inference cost. The findings offer a new perspective on controllable, interpretable personalization in LLMs through latent, sparsely routed parameters rather than external prompts or fine-tuning.

Abstract

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.

Your Language Model Secretly Contains Personality Subnetworks

TL;DR

This work demonstrates that diverse human-like personas are embedded as sparse, trainable subnetworks within pretrained LLMs. It introduces a training-free framework that uses activation-guided pruning to extract persona-specific subnetworks, plus a contrastive pruning variant to maximize separation between opposing personas. Empirical results across three datasets and multiple models show superior persona alignment over prompt-based and RAG baselines while maintaining general performance and reducing inference cost. The findings offer a new perspective on controllable, interpretable personalization in LLMs through latent, sparsely routed parameters rather than external prompts or fine-tuning.

Abstract

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.
Paper Structure (32 sections, 9 equations, 4 figures, 16 tables)

This paper contains 32 sections, 9 equations, 4 figures, 16 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Persona Modeling in LLMs
  4. Network Pruning and Sparse subnetworks
  5. Mechanistic Interpretability and Model Circuits
  6. Methods
  7. Problem Setup
  8. Persona subnetworks via Pruning
  9. Efficient Inference via Dynamic Masking
  10. Mask Analysis
  11. Experiments
  12. Experimental Settings
  13. Datasets
  14. Baselines
  15. Results and Analysis
  16. ...and 17 more sections

Figures (4)

  • Figure 1: The figure illustrates our pruning framework. Persona specific data is employed to compute activation statistics and importance scores, which are then used to rank parameters and construct masks that isolate persona-relevant subnetworks. Colored entries mark the Top-K parameters retained for each output neuron, while gray entries are pruned.
  • Figure 2: MBTI Heatmap.
  • Figure 3: Radar plots of MBTI task under different sparsity ratios for INFP, INFJ, INTP, and INTJ.
  • Figure 4: The results show distinct trends across sparsity values, where the x-axis denotes different sparsity levels.