aiqbench analysis
This research investigates the internal representation of human personas within large language models, revealing that distinct ideological traits are encoded primarily in the final third of decoder layers. The study highlights significant differences between how ethical perspectives and political ideologies occupy embedding spaces, offering critical insights into model interpretability and bias modulation. These findings advance our understanding of where specific value systems reside inside neural networks to guide future refinement efforts.
Localizing Persona Representations in LLMs
arXiv:2505.24539v4 Announce Type: replace-cross Abstract: We present a study on how and where personas -- defined by distinct sets of human characteristics, values, and beliefs -- are encoded in the representation space of large language models (LLMs). Using a range of dimension reduction and pattern recognition methods, we first identify the model layers that show the greatest divergence in encoding these representations. We then analyze the activations within a selected layer to examine how specific personas are encoded relative to others, including their shared and distinct embedding spaces. We find that, across multiple pre-trained decoder-only LLMs, the analyzed personas show large differences in representation space only within the final third of the decoder layers. We observe overlapping activations for specific ethical perspectives -- such as moral nihilism and utilitarianism -- suggesting a degree of polysemy. In contrast, political ideologies like conservatism and liberalism appear to be represented in more distinct regions. These findings help to improve our understanding of how LLMs internally represent information and can inform future efforts in refining the modulation of specific human traits in LLM outputs. Warning: This paper includes potentially offensive sample statements.
Abstract:We present a study on how and where personas -- defined by distinct sets of human characteristics, values, and beliefs -- are encoded in the representation space of large language models (LLMs). Using a range of dimension reduction and pattern recognition methods, we first identify the model layers that show the greatest divergence in encoding these representations. We then analyze the activations within a selected layer to examine how specific personas are encoded relative to others, including their shared and distinct embedding spaces. We find that, across multiple pre-trained decoder-only LLMs, the analyzed personas show large differences in representation space only within the final third of the decoder layers. We observe overlapping activations for specific ethical perspectives -- such as moral nihilism and utilitarianism -- suggesting a degree of polysemy. In contrast, political ideologies like conservatism and liberalism appear to be represented in more distinct regions. These findings help to improve our understanding of how LLMs internally represent information and can inform future efforts in refining the modulation of specific human traits in LLM outputs. Warning: This paper includes potentially offensive sample statements.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2505.24539 [cs.CL]
(or arXiv:2505.24539v4 [cs.CL] for this version)
https://doi.org/10.1609/aies.v8i1.36577 DOI(s) linking to related resources
Full text under the publisher's open licence · 204 words
Related coverage
[AINews] Opus 5.5 is good at explainer videos
Latent Space · Sep 29, 2026
[AINews] The Future of Latent Space
Latent Space · Sep 25, 2026
PersonaManifold: Revealing and Exploiting Curved Geometry in LLM Persona Representations
arXiv · Sep 29, 2026
Thousand-dimensional structure
AI Alignment Forum · Jul 30, 2026