Beyond the Hivemind: Escaping LLM Homogeneity via Meta-Persona Anchoring and Sequential Temperature Scaling

Researchers propose Meta-Persona Anchoring with Filtered Temperature Scaling (FTS) to reduce the 'Artificial Hivemind' effect in large language models, cutting average pairwise cosine similarity from approximately 0.85 to 0.65 on the INFINITY-CHAT dataset using open-weight models under 20B parameters. The method uses a two-stage generation process with self-selected personas and extreme temperature scaling (T ≥ 4.0) to increase output diversity.

arXiv:2608.02618v1 Announce Type: new Abstract: 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.