Facet-Level Persona Control by Trait-Activated Routing with Contrastive SAE for Role-Playing LLMs

View PDF HTML (experimental)

Abstract:Personality control in Role-Playing Agents (RPAs) is commonly achieved via training-free methods that inject persona descriptions and memory through prompts or retrieval-augmented generation, or via supervised fine-tuning (SFT) on persona-specific corpora. While SFT can be effective, it requires persona-labeled data and retraining for new roles, limiting flexibility. In contrast, prompt- and RAG-based signals are easy to apply but can be diluted in long dialogues, leading to drifting and sometimes inconsistent persona behavior. To address this, we propose a contrastive Sparse AutoEncoder (SAE) framework that learns facet-level personality control vectors aligned with the Big Five 30-facet model. A new 15,000-sample leakage-controlled corpus is constructed to provide balanced supervision for each facet. The learned vectors are integrated into the model's residual space and dynamically selected by a trait-activated routing module, enabling precise and interpretable personality steering. Experiments on Large Language Models (LLMs) show that the proposed method maintains stable character fidelity and output quality across contextualized settings, outperforming Contrastive Activation Addition (CAA) and prompt-only baselines. The combined SAE+Prompt configuration achieves the best overall performance, confirming that contrastively trained latent vectors can enhance persona control while preserving dialogue coherence. Dataset is available at: this https URL

Comments: Accepted in PAKDD 2026 special session on Data Science :Foundation and Applications

Subjects:

Computation and Language (cs.CL)

Cite as: arXiv:2602.19157 [cs.CL]

(or arXiv:2602.19157v2 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2602.19157

arXiv-issued DOI via DataCite

Submission history

From: Wenqiu Tang [view email] [v1] Sun, 22 Feb 2026 12:39:02 UTC (2,305 KB) [v2] Fri, 27 Mar 2026 01:56:36 UTC (2,305 KB)