PERSONA: Dynamic and Compositional Inference-Time Personality Control via Activation Vector Algebra

TL;DR

PERSONA introduces a training-free method for dynamic personality control in large language models by manipulating personality vectors in activation space, enabling fine-grained, compositional, and context-aware personality adjustments.

Contribution

It presents a novel framework that extracts and manipulates personality traits as orthogonal vectors in activation space, avoiding costly fine-tuning and enabling dynamic personality adaptation.

Findings

Achieves near-supervised performance on PersonalityBench without training.

Demonstrates effective dynamic personality adaptation with 91% win rates.

Shows that personality traits are mathematically tractable in model representations.

Abstract

Current methods for personality control in Large Language Models rely on static prompting or expensive fine-tuning, failing to capture the dynamic and compositional nature of human traits. We introduce PERSONA, a training-free framework that achieves fine-tuning level performance through direct manipulation of personality vectors in activation space. Our key insight is that personality traits appear as extractable, approximately orthogonal directions in the model's representation space that support algebraic operations. The framework operates through three stages: Persona-Base extracts orthogonal trait vectors via contrastive activation analysis; Persona-Algebra enables precise control through vector arithmetic (scalar multiplication for intensity, addition for composition, subtraction for suppression); and Persona-Flow achieves context-aware adaptation by dynamically composing these vectors during inference. On PersonalityBench, our approach achieves a mean score of 9.60, nearly matching the supervised fine-tuning upper bound of 9.61 without any gradient updates. On our proposed Persona-Evolve benchmark for dynamic personality adaptation, we achieve up to 91% win rates across diverse model families. These results provide evidence that aspects of LLM personality are mathematically tractable, opening new directions for interpretable and efficient behavioral control.