English

Multi-Personality Generation of LLMs at Decoding-time

Computation and Language 2026-01-16 v4 Artificial Intelligence

Abstract

Multi-personality generation for LLMs, enabling simultaneous embodiment of multiple personalization attributes, is a fundamental challenge. Existing retraining-based approaches are costly and poorly scalable, while decoding-time methods often rely on external models or heuristics, limiting flexibility and robustness. In this paper, we propose a novel Multi-Personality Generation (MPG) framework under the decoding-time combination paradigm. It flexibly controls multi-personality without relying on scarce multi-dimensional models or extra training, leveraging implicit density ratios in single-dimensional models as a "free lunch" to reformulate the task as sampling from a target strategy aggregating these ratios. To implement MPG efficiently, we design Speculative Chunk-level based Rejection sampling (SCR), which generates responses in chunks and parallelly validates them via estimated thresholds within a sliding window. This significantly reduces computational overhead while maintaining high-quality generation. Experiments on MBTI personality and Role-Playing demonstrate the effectiveness of MPG, showing improvements up to 16%-18%. Code and data are available at https://github.com/Libra117/MPG .

Keywords

Cite

@article{arxiv.2511.01891,
  title  = {Multi-Personality Generation of LLMs at Decoding-time},
  author = {Rongxin Chen and Yunfan Li and Yige Yuan and Bingbing Xu and Huawei Shen},
  journal= {arXiv preprint arXiv:2511.01891},
  year   = {2026}
}

Comments

Accepted by WSDM 2026

R2 v1 2026-07-01T07:19:53.853Z