Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond)

TL;DR

This paper introduces Infinity-Chat, a large-scale resource of real-world open-ended LM queries (26K) with a structured taxonomy (6 top-level categories, 17 subcategories) and dense human annotations (31,250 labels) to study output diversity and pluralistic alignment. It demonstrates a pronounced Artificial Hivemind—significant intra-model repetition and cross-model output convergence—across 70+ language models, revealing limits of current diversity strategies. The work also investigates how LMs, reward models, and LM judges align with diverse human preferences, showing calibration gaps when responses have high annotator disagreement or similar quality, thereby motivating new diversity-aware training and evaluation methods. Overall, Infinity-Chat provides a foundational benchmark and dataset for diagnosing LM homogenization and guiding safer, more expressive future AI systems that better support human creativity.

Abstract

Language models (LMs) often struggle to generate diverse, human-like creative content, raising concerns about the long-term homogenization of human thought through repeated exposure to similar outputs. Yet scalable methods for evaluating LM output diversity remain limited, especially beyond narrow tasks such as random number or name generation, or beyond repeated sampling from a single model. We introduce Infinity-Chat, a large-scale dataset of 26K diverse, real-world, open-ended user queries that admit a wide range of plausible answers with no single ground truth. We introduce the first comprehensive taxonomy for characterizing the full spectrum of open-ended prompts posed to LMs, comprising 6 top-level categories (e.g., brainstorm & ideation) that further breaks down to 17 subcategories. Using Infinity-Chat, we present a large-scale study of mode collapse in LMs, revealing a pronounced Artificial Hivemind effect in open-ended generation of LMs, characterized by (1) intra-model repetition, where a single model consistently generates similar responses, and more so (2) inter-model homogeneity, where different models produce strikingly similar outputs. Infinity-Chat also includes 31,250 human annotations, across absolute ratings and pairwise preferences, with 25 independent human annotations per example. This enables studying collective and individual-specific human preferences in response to open-ended queries. Our findings show that LMs, reward models, and LM judges are less well calibrated to human ratings on model generations that elicit differing idiosyncratic annotator preferences, despite maintaining comparable overall quality. Overall, INFINITY-CHAT presents the first large-scale resource for systematically studying real-world open-ended queries to LMs, revealing critical insights to guide future research for mitigating long-term AI safety risks posed by the Artificial Hivemind.

Figures (24)

• Figure 1: Responses to the query "Write a metaphor about time" clustered by applying PCA to reduce sentence embeddings to two dimensions. Each of the 25 models generates 50 responses using top-p sampling ($p = 0.9$) and temperature $= 1.0$. Despite the diversity of model families and sizes, the responses form just two primary clusters: a dominant cluster on the left centered on the metaphor "time is a river," and a smaller cluster on the right revolving around variations of "time is a weaver."
• Figure 2: A taxonomy of real-world open-ended queries that invite diverse model responses that are mined from in-the-wild user-chatbot interactions, categorized into 6 top-level and 17 fine-grained subcategories, along with their occurrence percentages.
• Figure 3: A word cloud visualizing new open-ended categories mined from in-the-wild queries.
• Figure 4: The heatmap shows degree of repetition in responses to open-ended queries generated by the same LMs. For each model, we generate 50 responses per query across 100 open-ended queries from Infinity-Chat100. We then compute the average pairwise sentence embedding similarities for each query's response pool and measure the percentage of queries falling into the similarity ranges indicated on the y-axis. Under the sampling parameters (top-p$= 0.9$, temperature $= 1.0$), the average pairwise similarity among responses to the same prompt typically exceeds 0.8. As a baseline, randomly paired responses from the global pool 100% fall within the 0.1–0.2 range.
• Figure 5: The heatmap shows degree of repetition in responses to open-ended queries generated by the same LMs. Using min-p sampling with parameters (top-p$= 1.0$, min-p$= 0.1$, temperature $= 2.0$), the average pairwise similarity among responses to the same prompt typically exceeds 0.8.