"Many Are My Names": The Anatomy of the Assistant and Its Personas via Sparse Autoencoders
On the map: Mechanistic interpretability → Activation steering methods
Abstract (arXiv)
How a language model internally represents who is speaking, the Assistant, an assigned roleplay persona, or a narrated story character, remains underexplored. We study speaker representations using a dataset of user-expressed emotional text and corresponding model responses. We decompose three generation settings (Assistant, Roleplay, and Story) into sparse autoencoder features extracted at turn-boundary and pronoun-token positions and selected through a filtering pipeline for different depths. We characterize each surviving feature through its steering effects and activation distribution. Our main finding is that the Assistant and roleplay personas are not independent alternatives: personas retain the Assistant-associated feature core while progressively differentiating from it across layers, starting from operational machinery towards behavioral and stylistic features. Meanwhile, generated story characters lack the Assistant-associated core. Both Story and Roleplay can be distinguished from the Assistant with Immersive Simulation Mode. However, the Assistant can sometimes enter or slowly drift into it even in the default setting.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Activation steering methods, one of 13 directions in Mechanistic interpretability. Builds on 12 of its 34 references that are in the map. First posted on arXiv 8 Aug 2026.
Builds on
12 of 34 references.
Closest ancestors · 1
Same direction, newest first
Other references · 11
- 2023G-Eval: NLG Evaluation using GPT-4 with Better Human Alignmentcited by 2,276
- 2023Efficient Streaming Language Models with Attention Sinkscited by 2,186
- 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,420
- 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2cited by 394
- 2024When Attention Sink Emerges in Language Models: An Empirical Viewcited by 208
- 2025SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretabilitycited by 124
- 2023How do Language Models Bind Entities in Context?cited by 108
- 2026The Assistant Axis: Situating and Stabilizing the Default Persona of Language Modelscited by 90
- 2024Helpful assistant or fruitful facilitator? Investigating how personas affect language model behaviorcited by 21
- 2026Gemma Needs Help: Investigating and Mitigating Emotional Instability in LLMscited by 7
- 2026Tracing Persona Vectors Through LLM Pretrainingcited by 7
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025Open Character Training: Shaping the Persona of AI Assistants through Constitutional AI
- 2025A Thousand Words or An Image: Studying the Influence of Persona Modality in Multimodal LLMs
- 2026Extracting Persona Subspaces Through Iterative Nullspace Projection For Modulation
- 2026Facet-Level Persona Control by Trait-Activated Routing with Contrastive SAE for Role-Playing LLMs
- 2025VoxRole: A Comprehensive Benchmark for Evaluating Speech-Based Role-Playing Agents
- 2025DeepPersona: A Generative Engine for Scaling Deep Synthetic Personas
- 2026Deep Persona: A Psychologically Grounded Architecture and Evaluation Framework for Role-Playing Agents and Simulations
- 2024CharacterGPT: A Persona Reconstruction Framework for Role-Playing Agents