Output Diversity in Language Models

Follow this topic Ask about this topic

Latest papers

38 papers9 papers in the last four weeks, up 80% on the four weeks before. 0.1% of all new papers.

Richness Is Not Evenness: Ordered Renyi Crossings under Self-Training
Nov 6, 2026Pranjal SrivastavaSelf-Training for Language Models Output Diversity in Language Models
University of Michigan
Test-Time Scaling for Open-Ended Reasoning at Small Model Scale
Nov 1, 2026Adam Y. Ahmed, Abdullah G. ElafifiTest-Time Scaling Output Diversity in Language Models
Revisiting LLM Homogeneity Through a Multidimensional Lens
Nov 1, 2026Ziliang Qiu, Renfen Hu, Yi Ding +1Homogenization in Language Models Output Diversity in Language Models
University of Illinois Urbana-Champaign · Beijing Normal University · Purdue University
Character-Name Concentration in LLM Fiction: Measuring Displacement and Mitigating Repetition
Oct 31, 2026Alejandro Mozo Quesada, Sergio Gálvez Rojas, Francisco López ValverdeRepetition in Language Generation Controllable Text Generation
Universidad de Málaga, Spain
A Russian, a Brazilian, and a German Walk into a Bar, Ask the Same Question-- And Get Three Different Answers: Profiling LLM Perspective Diversity on Contested Topics
Oct 25, 2026Wiem Ben Rim, Yao Lu, Pontus Stenetorp +1Cross-Cultural Language Model Evaluation Output Diversity in Language Models
University College London
PRISM: Promoting Sampling Diversity of Large Language Models with Implicit Stochastic Modulation
Oct 23, 2026Qixin Xu, Zihan Zhu, Xiaojiang Liu +1Output Diversity in Language Models LLM Training
Tsinghua University. · Apple.
MemBubble: Benchmarking How Agent Memory Shapes LLM Response Diversity
Oct 18, 2026Jiajun Tan, Fei Sun, Xinyu Ma +6LLM Agent Memory Output Diversity in Language Models
State Key Laboratory of AI Safety, Institute of Computing Technology, CAS · University of Chinese Academy of Sciences · Baidu Inc.
Language Models are Lazy Poets: Creative Mode Collapse at Scale
Oct 16, 2026Jacob A. Matthews, Sil HamiltonOutput Diversity in Language Models Mode Collapse
Cornell University
Analyzing Occupational Gender Bias across the LLM Training Lifecycle
Oct 16, 2026Manthan Gadhia, Alexander Hoyle, Karolina StańczakGender Bias in Language Models Language Model Post-Training
Department of Humanities, Social, and Political Sciences, ETH Zürich · TU Wien, Complexity Science Hub · ETH AI Center, ETH Zürich
The Multilingual Artificial Hivemind: Do All LLMs Think Alike in Every Language?
Oct 12, 2026Xaver Raphael Davey, Shubhra Mishra, Anton Adelöw +1Output Diversity in Language Models Homogenization in Language Models
KTH Royal Institute of Technology
Gacha Decoding: Eliciting Diverse Generations Through Instruction Following
Oct 1, 2026Scott Geng, Yufei Zhang, Joseph Lee +3Language Model Decoding Diverse Text Generation
University of Washington · Meta Superintelligence Labs
How Strong Is the Evidence for the Artificial Hivemind? Reevaluating Evidence for the Open-Ended Homogeneity of Language Models
Sep 27, 2026Rylan Schaeffer, Brando Miranda, Joshua Kazdan +2Output Diversity in Language Models Homogenization in Language Models
Stanford Computer Science
Improving the Diversity of LLM Outputs without a Trade-off
Sep 27, 2026Ryoma SatoOutput Diversity in Language Models Language Model Decoding
National Institute of Informatics
Breaking Homogeneity: Diversifying Persona Sets for Creative LLM Outputs
Sep 24, 2026Sang Bin Moon, Nicole Cho, Daniel Borrajo +2Output Diversity in Language Models Diverse Text Generation
School of Electrical and Computer Engineering Purdue University West Lafayette, IN 47909, USA · J.P. Morgan AI Research New York, NY 10172, USA
Fine-Tuning Fixes Mode Collapse and Over-Dispersion in LLMs
Sep 15, 2026Kirill Skobelev, Eric Fithian, X. Y. HanOutput Diversity in Language Models Mode Collapse
Northwestern University · University of Chicago
Evaluating Multiple LLM Generations with Validated Task Coverage
Aug 25, 2026Florian Le Bronnec, Rio YokotaOutput Diversity in Language Models LLM Evaluation
RIKEN Center for Computational Science, Tokyo, Japan
CreativeInstruct: Scalably Teaching LLMs to Balance Quality, Creativity, and Diversity
Aug 7, 2026Ananya Sahu, Mohit Bansal, Elias Stengel-EskinInstruction Tuning Diverse Text Generation
Columbia · UNC Chapel Hill · University of Texas at Austin
Beyond "AI Language": The case for the idiolectal nature of LLM output
Aug 6, 2026Karolina Rudnicka, Thomas Stephan JuzekLanguage Model Fingerprinting Output Diversity in Language Models
Faculty of Languages, University of Gdańsk, Poland · Department of Scientific Computing, Florida State University, United States
On the Diversity of Analogy Making in Large Language Models
Aug 4, 2026Yuanhao Shen, Daniel Xavier de Sousa, Caio César Sifuentes Barcelos +2LLM Evaluation Analogical Reasoning for Language Models
Department of Electrical and Computer Engineering & Ingenuity Labs Research Institute, Queen’s University, Canada · Instituto Federal de Goiás, Anápolis, Brazil · National Research Council Canada.
Structured Output Collapses Answer Diversity Across 44 Language Models
Jul 20, 2026Tapan ParikhLLM Evaluation Structured Prompting
Cornell Tech
The One-Word Census: Answer-Choice Conformity Across 105 Language Models
Jul 14, 2026Tapan ParikhLLM Evaluation Language Model Generation Evaluation
Cornell Tech
Where You Inject Diversity Matters: A Unified Framework for Diverse Generation
Jun 9, 2026Cheng Zhang, Rui Xin, Chudi ZhongOpen-Ended Generation Controllable Text Generation
UNC Chapel Hill · University of Washington
"I've Seen How This Goes": Characterizing Diversity via Progressive Conditional Surprise
Jun 1, 2026Matthew Khoriaty, David Williams-King, Shi FengDiverse Text Generation Generative AI Evaluation
ERA Fellowship · George Washington University
Fine-Tuned Language Models Exhibit Improved Information Conveyance
May 29, 2026Yuwei Cheng, Weiyi Tian, Haifeng XuLanguage Model Generation Evaluation LLM Fine-Tuning
Department of Statistics, University of Chicago, Chicago, IL 60637 · Department of Data Science, University of Chicago, Chicago, IL 60637 · Department of Computer Science, University of Chicago, Chicago, IL 60637
Recovering Diversity Without Losing Alignment: A DPO Recipe for Post-Trained LLMs
May 28, 2026Vinay Samuel, Yapei Chang, Mohit IyyerLLM Alignment Direct Preference Optimization
University of Maryland, College Park
Semantic Flow Regularization: Teaching LLMs to Generate Diverse Yet Coherent Responses
May 27, 2026Kerui Peng, Feifei Li, Xingyu Fan +1Controllable Text Generation Diverse Text Generation
WeChat, Tencent Inc., Beijing, China
Sampling More, Getting Less: Calibration is the Diversity Bottleneck in LLMs
May 11, 2026Amin Banayeeanzade, Qingchuan Yang, Dhruv Tarsadiya +6LLM Evaluation Language Model Decoding
♠University of Southern California · ♣Capital One
Large Language Models Explore by Latent Distilling
Apr 27, 2026Yuanhao Zeng, Ao Lu, Lufei Li +3Language Model Decoding Test-Time Training
School of Information Science and Technology, ShanghaiTech University, Shanghai, China · State Key Laboratory of General Artificial Intelligence, BIGAI, Beijing, China.
Beyond One Output: Visualizing and Comparing Distributions of Language Model Generations
Apr 20, 2026Emily Reif, Claire Yang, Jared Hwang +3Open-Ended Generation Data Visualization
University of Washington, Seattle, WA, USA · Allen Institute for AI
Where does output diversity collapse in post-training?
Apr 17, 2026Constantinos Karouzos, Xingwei Tan, Nikolaos AletrasOutput Diversity in Language Models Language Model Post-Training
School of Computer Science University of Sheffield, UK

Nov 6, 2026J/K move · Enter open · S save

Richness Is Not Evenness: Ordered Renyi Crossings under Self-Training

Pranjal Srivastava

University of Michigan

Self-Training for Language Models Output Diversity in Language Models Model Collapse

Open paper AskPDF arXiv

Recursive self-training of language models, a setting studied in model collapse, can reduce vocabulary while making frequent words, mostly function words, less dominant. We fine-tune English GPT-2 on its own output for nine generations, vary the training-data sampling temperature, and measure every model with one fixed sampler. As training temperature rises, Renyi entropy changes turn from loss to gain at distinct temperatures: min-entropy first, then collision entropy, Shannon entropy and observed richness. Between the outer crossings, word types fall while the most-common word becomes less dominant. Untruncated training at T = 1 lies in this interval: at 13,000 words, generation 9 has 24% fewer types than generation 0 yet collision entropy is 0.91 bits higher. The ordering persists with a finer grid, different measurement samplers, a fixed word budget and five fresh training seeds. Min-entropy again crosses before richness in four further settings (in Hindi only at interpolated roots) and in 12 of 13 distinct measurements under eight text representations. Next-token identities and a categorical simulation isolate decoding and sampling effects; the crossing order remains empirical. Fine-tuning on human text can produce the same split while likelihood improves, and no single order's sign reliably identifies worse likelihood across the tested settings. We recommend reporting entropy profiles at stated budgets with an independent quality measure.