发表机构
Politecnico di Milano; Information Processing and Telecommunications Center; Universidad Politécnica de Madrid; Banco de España（米兰理工大学; 信息处理与电信中心; 马德里理工大学; 西班牙银行）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对大语言模型的同质化问题，提出元角色锚定结合过滤式温度缩放的框架，在INFINITY-CHAT数据集的约200亿参数模型上，将平均成对余弦相似度从0.85降至0.65，缩小了人工模式坍缩与人类多样性的差距并开源了框架。
AI 中文摘要
近期研究在大语言模型（LLMs）中发现了一种“人工群体思维”效应，该效应会导致模型即使面对开放性问题也收敛于狭窄、同质化的共识。这种语义坍缩限制了AI的多样性，即便在高温采样下，响应间的相似度仍高达约0.80-0.90。本文提出一种新的缓解框架以提升多样性：元角色锚定（Meta-Persona Anchoring）结合过滤式温度缩放（Filtered Temperature Scaling, FTS）。该方法采用两阶段生成流程：首先，提示模型自行选择独特、具个人特质的角色以锚定其生成起点；其次，应用双阶段采样筛选器，先通过Top-p过滤保留语法有效性，再对留存候选应用极端温度缩放（T≥4.0）以探索更宽泛的概率分布。我们在INFINITY-CHAT数据集上，对约200亿参数的最先进开源权重模型评估该方法。结果显示语义收敛性显著降低，平均成对余弦相似度从约0.85降至约0.65。我们的方案使多数问题的相似度低于0.7阈值，有效缩小了人工模式坍缩与人类层面类型多样性的差距。我们将实现作为开源框架发布，以支持更多样、更具创造性的AI部署。
英文摘要
Recent studies have identified an ``Artificial Hivemind'' effect in Large Language Models (LLMs) causing models to converge on a narrow, homogenized consensus even for open questions. This semantic collapse limits the diversity of AI, resulting in high inter-response similarity ($\approx 0.80-0.90$) even under high-temperature sampling. In this paper, we propose a novel mitigation framework to increase diversity: Meta-Persona Anchoring combined with Filtered Temperature Scaling (FTS). Our approach utilizes a two-stage generation process: first, the model is prompted to self-select a unique, idiosyncratic persona to anchor its starting point; second, we apply a dual-stage sampling sieve, utilizing Top-$p$ filtering to preserve grammatical validity followed by extreme temperature scaling ($T \ge 4.0$) on the surviving candidates to explore the broadened probability distribution. We evaluate our method using the INFINITY-CHAT dataset on state-of-the-art open weight models under $\sim$20B parameters. Our results demonstrate a significant reduction in semantic convergence, with average pairwise cosine similarity dropping from ($\approx 0.85$) to ($\approx 0.65$). Our scheme achieves a majority of questions below the 0.7 threshold, effectively reducing the gap between artificial mode collapse and human-level typological diversity. We provide our implementation as an open-source framework to enable more diverse and creative AI deployments.
