PERSONA：通过激活向量代数实现动态和组合性的人格控制
PERSONA: Dynamic and Compositional Inference-Time Personality Control via Activation Vector Algebra
AI总结：
PERSONA通过激活向量代数实现动态人格控制，无需训练即可达到微调水平性能，并在多个基准测试中表现出色。
摘要：
当前大型语言模型中的人格控制方法依赖于静态提示或昂贵的微调，无法捕捉人类特质的动态和组合性质。我们引入PERSONA，一种无需训练的框架，通过直接操纵激活空间中的人格向量实现微调级别的性能。我们的关键见解是，人格特质出现在模型表示空间中可提取的、近似正交的方向上，这些方向支持代数运算。该框架通过三个阶段运作：Persona-Base通过对比激活分析提取正交特质向量；Persona-Algebra通过向量运算（标量乘法用于强度，加法用于组合，减法用于抑制）实现精确控制；Persona-Flow通过在推理过程中动态组合这些向量实现上下文感知的适应。在PersonalityBench上，我们的方法实现了9.60的平均分数，几乎达到了监督微调上限9.61，而无需任何梯度更新。在我们提出的Persona-Evolve基准测试中，我们实现了在不同模型家族中高达91%的胜利率。这些结果提供了证据，表明LLM人格的某些方面在数学上是可处理的，为可解释和高效的可控行为开辟了新方向。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
