EMNLP Findings2025
Dissecting Persona-Driven Reasoning in Language Models via Activation Patching
Abstract
Large language models (LLMs) exhibit remarkable versatility in adopting diverse personas. In this study, we examine how assigning a persona influences a model's reasoning on an objective task. Using activation patching, we take a first step toward understanding how key components of the model encode persona-specific information. Our findings reveal that the early Multi-Layer Perceptron (MLP) layers attend not only to the syntactic structure of the input but also process its semantic content. These layers transform persona tokens into richer representations, which are then used by the middle Multi-Head Attention (MHA) layers to shape the model's output. Additionally, we identify specific attention heads that disproportionately attend to racial and color-based identities. 1
Ask Lune
Know this paper in minutes, not an afternoon.
Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.
1 top-tier paper has cited it since. Know what they found before your reviewers do.
Lune can draw on
the full paper · the top-tier papers that cite it · related work
Free to start. No credit card required.
Lineage
- Locating and Editing Factual Associations in GPT cites Investigating Gender Bias in Language Models Using Causal Mediation Analysis
- Towards Best Practices of Activation Patching in Language Models: Metrics and Methods cites Locating and Editing Factual Associations in GPT
- Towards Best Practices of Activation Patching in Language Models: Metrics and Methods cites Investigating Gender Bias in Language Models Using Causal Mediation Analysis
- In-Context Impersonation Reveals Large Language Models' Strengths and Biases cites Measuring Massive Multitask Language Understanding
Cited by top-tier papers1
Builds on13
- Measuring Massive Multitask Language UnderstandingDan Hendrycks, Collin Burns, Steven Basart, Andy Zou et al.ICLR 2021 · 7,905 citations
- Locating and Editing Factual Associations in GPTKevin Meng, David Bau, Alex Andonian, Yonatan BelinkovNeurIPS 2022 · 3,415 citations
- Investigating Gender Bias in Language Models Using Causal Mediation AnalysisJesse Vig, Sebastian Gehrmann, Yonatan Belinkov, Sharon Qian et al.NeurIPS 2020 · 851 citations
- In-Context Impersonation Reveals Large Language Models' Strengths and BiasesLeonard Salewski, Stephan Alaniz, Isabel Rio-Torto, Eric Schulz et al.NeurIPS 2023 · 259 citations
- Towards Best Practices of Activation Patching in Language Models: Metrics and MethodsFred Zhang, Neel NandaICLR 2024 · 233 citations
Related papers
- Bias Runs Deep: Implicit Reasoning Biases in Persona-Assigned LLMsShashank Gupta, Vaishnavi Shrivastava, Ameet Deshpande, Ashwin Kalyan et al.ICLR 2024 · 212 citations
- Persona-Assigned Large Language Models Exhibit Human-Like Motivated ReasoningSaloni Dash, Amélie Reymond, Emma S. Spiro, Aylin CaliskanACL Findings 2026 · 1 citation
- Quantifying the Persona Effect in LLM SimulationsTiancheng Hu, Nigel CollierACL 2024 · 22 citations
- Your Language Model Secretly Contains Personality SubnetworksRuimeng Ye, Zihan Wang, Zinan Ling, Yang Xiao et al.ICLR 2026 · 4 citations
- Conceptual Hierarchies within LLMsTiago Almeida, Zining Zhu, Yue NingACL Findings 2026
