解决对话中角色漂移问题，让大模型长期保持一致人设。
SPASM: Stable Persona-driven Agent Simulation for Multi-turn Dialogue Generation
- 用框架拆解对话生成，先建可信人设再对话
- 引入视角无关表示，显著减少人设漂移和模仿现象
- 适合训练/评估需要稳定角色的对话系统
大型语言模型在多轮对话场景（如辅导、支持、咨询）中的可靠性依赖于长期保持一致的角色、人格和目标。当用于生成合成对话以训练或评估时，模型间对话易出现身份错误，如人设漂移、角色混淆和‘回声’现象（一方逐渐模仿对方）。本文提出SPASM（Stable Persona-driven Agent Simulation for Multi-turn dialogue generation），一种模块化、以稳定性为先的框架，将模拟过程分解为：(i) 通过模式采样、合理性验证和自然语言构建人设；(ii) 客户-回应者对话生成；(iii) 一致性终止检测。为在不修改模型权重的前提下提升长程稳定性，提出共情上下文投影（ECP）：对话历史以视角无关形式存储，并在生成前确定性地投影到每个代理的主观视角。在三种LLM主干（GPT-4o-mini、DeepSeek-V3.2、Qwen-Plus）和九组客户-回应者配对下，构建了包含4,500个人设和45,000段对话的数据集（每组500个人设×每人人设10段对话）。消融实验表明，ECP显著降低人设漂移，在人工评估中彻底消除回声；嵌入分析恢复了人设结构并揭示强烈的回应者驱动交互几何。代码已开源。
原文摘要 · Abstract (English)
Large language models are increasingly deployed in multi-turn settings such as tutoring, support, and counseling, where reliability depends on preserving consistent roles, personas, and goals across long horizons. This requirement becomes critical when LLMs are used to generate synthetic dialogues for training and evaluation, since LLM--LLM conversations can accumulate identity-related failures such as persona drift, role confusion, and "echoing", where one agent gradually mirrors its partner. We introduce SPASM (Stable Persona-driven Agent Simulation for Multi-turn dialogue generation), a modular, stability-first framework that decomposes simulation into (i) persona creation via schema sampling, plausibility validation, and natural-language persona crafting, (ii) Client--Responder dialogue generation, and (iii) termination detection for coherent stopping. To improve long-horizon stability without changing model weights, we propose Egocentric Context Projection (ECP): dialogue history is stored in a perspective-agnostic representation and deterministically projected into each agent's egocentric view before generation. Across three LLM backbones (GPT-4o-mini, DeepSeek-V3.2, Qwen-Plus) and nine Client--Responder pairings, we construct a dataset of 4,500 personas and 45,000 conversations (500 personas X 10 conversations per pairing). Ablations show ECP substantially reduces persona drift and, under human validation, eliminates echoing; embedding analyses recover persona structure and reveal strong responder-driven interaction geometry. Our code is available at https://github.com/lhannnn/SPASM.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。