LLMs show swarm intelligence potential, reducing errors by 37%

A new research paper explores the potential of large language models (LLMs) to replicate the accuracy of human swarm intelligence. The study involved 960 prompts across GPT-5, Gemini 2.5 Pro, and Claude Sonnet 4.5, demonstrating that aggregating responses from these models consistently reduced errors by up to 37 percentage points. The research also found that LLMs exhibit a degree of metacognitive awareness, correlating confidence intervals with estimation errors, suggesting their utility in organizational decision-making. AI

IMPACT Demonstrates LLMs can aggregate information effectively, potentially improving AI-assisted decision-making in organizations.

RANK_REASON Research paper published on arXiv detailing LLM capabilities. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Claude Sonnet 4.5
• CORE Recommender
• DagsHub
• Gemini 2.5 Pro
• Gotit.pub
• GPT-5
• Hugging Face
• Influence Flower
• ScienceCast
• Spearman's ho

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →