What and Whose Knowledge? Measuring Epistemic Diversity in Large Language Models

arXiv:2510.04226 · cs.CL, cs.AI, cs.CY, cs.IR, cs.LG · Submitted 2025-10-05 · Read on arXiv

cs.CL, cs.AI, cs.CY, cs.IR, cs.LG

Submitted: 2025-10-05

Updated: 2026-08-31

Comments: Accepted to EMNLP 2026; 18 pages, 8 figures, 4 tables; v2: Fixed the modeling for table 3, random effect is the model version; v3: Fixed minor formatting issues in tables 2 and 3; v4: Fixed some typos and model description; v5: Updated metadata; v6: Improved search baseline, writing revisions, added comparisons to semantic similarity only approaches; v7 changelog: EMNLP camera ready

Code: https://github.com/dwright37/llm-knowledge

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• RAG — 51× in this paper · explained in 3 episodes
• Semantic Similarity — 14× in this paper · explained in 2 episodes
• Clustering — 12× in this paper · explained in 2 episodes
• Cosine Similarity — 9× in this paper · explained in 3 episodes
• Embeddings — 7× in this paper · explained in 2 episodes
• Homogenization — 4× in this paper · explained in 2 episodes
• LLM-as-a-Judge — 4× in this paper · explained in 3 episodes
• Large Language Models (LLMs) — 4× in this paper · explained in 8 episodes
• Alignment — 3× in this paper
• Benchmarking — 3× in this paper · explained in 4 episodes
• Abundance — 2× in this paper · explained in 1 episode
• Accuracy — 2× in this paper · explained in 1 episode
• Core — 2× in this paper · explained in 2 episodes
• Empirical Methods in Natural Language Processing — 2× in this paper
• Generalizability — 2× in this paper · explained in 6 episodes
• Homogeneity — 2× in this paper · explained in 2 episodes
• Retrieval-Augmented Generation (RAG) — 2× in this paper · explained in 18 episodes
• Accountability — 1× in this paper · explained in 2 episodes
• Adam — 1× in this paper · explained in 2 episodes
• Benchmark — 1× in this paper · explained in 2 episodes
• Cooperation — 1× in this paper · explained in 2 episodes
• Downstream Tasks — 1× in this paper · explained in 2 episodes
• F1 Score — 1× in this paper · explained in 3 episodes
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Hallucination — 1× in this paper · explained in 4 episodes
• Transformer — 1× in this paper · explained in 2 episodes

Sources

• Towards Measuring the Representation of Subjective Global Opinions in Language Models
• A survey of diversity quantification in natural language processing: The why, what, where and how
• Whose Name Comes Up? Auditing LLM-Based Scholar Recommendations
• Poor Alignment and Steerability of Large Language Models: Evidence from College Admission Essays
• How much do language models memorize?
• Bias patterns in the application of LLMs for clinical decision support: A comprehensive study
• The Lock-in Hypothesis: Stagnation by Algorithm
• GRADE: Quantifying Sample Diversity in Text-to-Image Models
• IssueBench: Millions of Realistic Prompts for Measuring Issue Bias in LLM Writing Assistance
• The Shrinking Landscape of Linguistic Diversity in the Age of Large Language Models
• The Homogenizing Effect of Large Language Models on Human Expression and Thought
• We're Different, We're the Same: Creative Homogeneity Across LLMs
• Echoes in AI: Quantifying lack of plot diversity in LLM outputs

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering