arXiv 2510.178442025-10-10cs.CLcs.AIcs.MA

使用多智能体大型语言模型建模分层意识

Modeling Layered Consciousness with Multi-Agent Large Language Models

Samsung Electronics（三星电子）；Hanyang University（汉阳大学）；Miami University（迈阿密大学）；The Catholic University of Korea College of Medicine（韩国天主教大学医学院）；Artificial Consciousness Lab, MODULABS（MODULABS 人工意识实验室）

Sang Hun Kim, Jongmin Lee, Dongkyu Park, So Young Lee, Yosep Chong

备注：: 20 pages, 4 figures, accepted for presentation at EMNLP 2025 Workshop on Active and Passive LLM Personalization (PALS) OpenReview: https://openreview.net/forum?id=rUtNkYvGJI

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文提出受精神分析理论启发的多智能体 Psychodynamic Model，通过个性化模块和参数高效微调模拟分层意识，实验表明其情感表现更稳定且更受偏好。

摘要：

我们提出一个基于精神分析理论的多智能体框架，用于在大型语言模型（LLM）中建模人工意识。我们的 Psychodynamic Model（心理动力学模型）通过智能体交互模拟自我意识、前意识和无意识，并由结合固定特质与动态需求的 Personalization Module（个性化模块）引导。该系统在情感丰富的对话上采用参数高效微调，并在八种个性化条件下接受评估。以 LLM 作为评判者的方法显示，微调模型获得 71.2% 的偏好，同时情感深度提升、输出方差降低，证明了其在适应性、个性化认知方面的潜力。

英文摘要：

We propose a multi-agent framework for modeling artificial consciousness in large language models (LLMs), grounded in psychoanalytic theory. Our \textbf{Psychodynamic Model} simulates self-awareness, preconsciousness, and unconsciousness through agent interaction, guided by a Personalization Module combining fixed traits and dynamic needs. Using parameter-efficient fine-tuning on emotionally rich dialogues, the system was evaluated across eight personalized conditions. An LLM as a judge approach showed a 71.2\% preference for the fine-tuned model, with improved emotional depth and reduced output variance, demonstrating its potential for adaptive, personalized cognition.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。