The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language?
Across 23 open-weight models and 13 languages, we find that the artificial hivemind effect is pervasive but uneven in its extent.
ProofPaper ↗
Key points
- Text generated by Large Language Models (LLMs) often remains limited in diversity compared to that produced by humans.
- Recent advances demonstrate the presence of inter- and intra-model collapse in open-ended generation of English text, raising concerns about the long-term homogenization of LLM, and more importantly, human thought.
- In this work, we investigate the role of language on output diversity.
- Specifically, we aim to understand whether multilingual LLM-generated text is also homogeneous, and whether this homogeneity matches what we observe in humans.
Sources (1)
- [1]The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language?arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 12, 01:20 PM
Across 23 open-weight models and 13 languages, we find that the artificial hivemind effect is pervasive but uneven in its extent.
Text generated by Large Language Models (LLMs) often remains limited in diversity compared to that produced by humans.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 12, 2026Autoregressive Decoding as a Martingale: Characterizing the Stochastic Dynamics of LLM Generation
- Oct 12, 2026Omni-Modal LLMs Know More Than They Say
- Oct 10, 2026SocraticRepair: Planning-Guided Socratic Tutoring for Program Debugging and Misconception Repair
- Oct 10, 2026Beyond Adjacent Layers: Graph-Guided Layer Fusion for Compressing Large Language Models
- Oct 10, 2026Toward System One Clinical Intelligence: Parameter-Efficient Fine-Tuning and Counterfactual Explainability of Decision Foundation Models
- Oct 10, 2026SAGE: Pushing 2-Bit Vector-Quantized LLMs toward Real-World Usability