BILLY: Steering Large Language Models via Merging Persona Vectors for Creative Generation

2025/10/11 by Pai, Tsung-Min, J Wang, Wang, Jui-I +8
Computer Science · Medicine · #Artificial Intelligence (cs.AI) #Artificial Intelligence in Healthcare and Education #Computation and Language (cs.CL) #FOS: Computer and information sciences #Persona Design and Applications #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2510.10157

openalex publication_date 2025/10/11 · openalex created_date 2025/10/15 · openalex updated_date 2026/07/28

Abstract

Multi-LLM systems enhance the creativity of large language models by simulating human collective intelligence but suffer from significant drawbacks, such as high computational costs and inference latency. To address these limitations, we propose BILLY (BlendIng persona vectors for Large Language model creativitY), a training-free framework that captures the benefits of multi-LLM collaboration, i.e. inducing diverse perspectives and specialized expertise, within a single model. BILLY operates by extracting and blending multiple distinct persona vectors directly in the model's activation space. We steer the model's generation process with this merged vector while inference, enabling multi-perspective output without explicit multi-LLM communication. Our experiments across creativity-oriented benchmarks demonstrate that BILLY surpasses single model prompting and traditional multi-LLM approaches, while substantially reducing inference time and computational costs. Our analyses further reveal that distinct persona vectors can be blended to achieve both effective control over complementary aspects of generation and greater interpretability.

Citations

• Personality Vector: Modulating Personality of Large Language Models by Model Merging
• Rethinking Creativity Evaluation: A Critical Analysis of Existing Creativity Evaluations
• Creativity in LLM-based Multi-Agent Systems: A Survey
• Adaptive Helpfulness-Harmlessness Alignment with Preference Vectors
• Why Do Multi-Agent LLM Systems Fail?
• Personalized Text Generation with Contrastive Activation Steering
• Divergent Thoughts toward One Goal: LLM-based Multi-Agent Collaboration System for Electronic Design Automation
• AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders
• Multi-Agent Collaboration Mechanisms: A Survey of LLMs
• SimTube: Generating Simulated Video Comments through Multimodal AI and User Personas
• Many Heads Are Better Than One: Improved Scientific Idea Generation by A LLM-Based Multi-Agent System
• Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers
• HoLLMwood: Unleashing the Creativity of Large Language Models in Screenwriting via Role Playing
• Personalized Steering of Large Language Models: Versatile Steering Vectors Through Bi-directional Preference Optimization
• Synthesizing Programmatic Reinforcement Learning Policies with Large Language Model Guided Search
• LLM Discussion: Enhancing the Creativity of Large Language Models via Discussion Framework and Role-Play
• From Persona to Personalization: A Survey on Role-Playing Language Agents
• Steering Llama 2 via Contrastive Activation Addition
• Improving Activation Steering in Language Models with Mean-Centring
• Personalized Soups: Personalized Large Language Model Alignment via Post-hoc Parameter Merging
• Bootstrap Your Own Skills: Learning to Solve New Tasks with Large Language Model Guidance
• A Confederacy of Models: a Comprehensive Evaluation of LLMs on Creative Writing
• Corex: Pushing the Boundaries of Complex Reasoning through Multi-Model Collaboration
• Art or Artifice? Large Language Models and the False Promise of Creativity
• Large Language Models for Automated Open-domain Scientific Hypotheses Discovery
• Steering Language Models With Activation Engineering
• AIKernel Semantic DSL Compiler and Deterministic Agent Execution Architecture
• ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debate
• Unleashing the Emergent Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration
• Inference-Time Intervention: Eliciting Truthful Answers from a Language Model
• Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate
• Improving Factuality and Reasoning in Language Models through Multiagent Debate
• Resource allocation in dynamic multiagent systems
• A Dynamic LLM-Powered Agent Network for Task-Oriented Agent Collaboration

Related

• Multi-Novelty: Improve the Diversity and Novelty of Contents Generated by Large Language Models via inference-time Multi-Views Brainstorming
• User-Controlled Knowledge Fusion in Large Language Models: Balancing Creativity and Hallucination
• Scaffolding Creativity: How Divergent and Convergent LLM Personas Shape Human Machine Creative Problem-Solving
• A Mathematical Abstraction for Balancing the Trade-off Between Creativity and Reality in Large Language Models
• Divergent Creativity in Humans and Large Language Models