mlsanityData to 14 Oct 2026

Effects of diversity incentives on sample diversity and downstream model performance in LLM-based text augmentation

arXiv 2401.06643 ↗First posted 12 Jan 2024cs.CLcited by 17

Not on the map. Why?

Abstract (arXiv)

The latest generative large language models (LLMs) have found their application in data augmentation tasks, where small numbers of text samples are LLM-paraphrased and then used to fine-tune downstream models. However, more research is needed to assess how different prompts, seed data selection strategies, filtering methods, or model settings affect the quality of paraphrased data (and downstream models). In this study, we investigate three text diversity incentive methods well established in crowdsourcing: taboo words, hints by previous outlier solutions, and chaining on previous outlier solutions. Using these incentive methods as part of instructions to LLMs augmenting text datasets, we measure their effects on generated texts lexical diversity and downstream model performance. We compare the effects over 5 different LLMs, 6 datasets and 2 downstream models. We show that diversity is most increased by taboo words, but downstream model performance is highest with hints.

Key facts. Cited by 17 papers in the map (data through 14 Oct 2026). Builds on 28 of its 46 references that are in the map. First posted on arXiv 12 Jan 2024.

Builds on

28 of 46 references.

References · 28

  1. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  2. 2019BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehensioncited by 5,893
  3. 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,418
  4. 2015Character-level Convolutional Networks for Text Classificationcited by 2,330
  5. 2020Making Pre-trained Language Models Better Few-shot Learnerscited by 1,256
  6. 2019Quantifying the Carbon Emissions of Machine Learningcited by 521
  7. 2020On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselinescited by 255
  8. 2020Reformulating Unsupervised Style Transfer as Paraphrase Generationcited by 183
  9. 2023Platypus: Quick, Cheap, and Powerful Refinement of LLMscited by 150
  10. 2023Few-shot Fine-tuning vs. In-context Learning: A Fair Comparison and Evaluationcited by 136
  11. 2018Cross-Lingual Transfer Learning for Multilingual Task Oriented Dialogcited by 125
  12. 2023Finding Support Examples for In-Context Learningcited by 104
16 more
  1. 2023Task Contamination: Language Models May Not Be Few-Shot Anymorecited by 99
  2. 2022Data Curation Alone Can Stabilize In-context Learningcited by 59
  3. 2020Neural Syntactic Preordering for Controlled Paraphrase Generationcited by 56
  4. 2021An Investigation of the (In)effectiveness of Counterfactually Augmented Datacited by 42
  5. 2023ChatGPT to Replace Crowdsourcing of Paraphrases for Intent Classification: Higher Diversity and Comparable Model Robustnesscited by 41
  6. 2020Paraphrase Generation as Zero-Shot Multilingual Translation: Disentangling Semantic Similarity from Lexical and Syntactic Diversitycited by 35
  7. 2023ZeroShotDataAug: Generating and Augmenting Training Data with ChatGPTcited by 32
  8. 2019Outlier Detection for Improved Data Quality and Diversity in Dialog Systemscited by 28
  9. 2022Novelty Controlled Paraphrase Generation with Retrieval Augmented Conditional Prompt Tuningcited by 25
  10. 2021Directed Diversity: Leveraging Language Embedding Distances for Collective Creativity in Crowd Ideationcited by 23
  11. 2023A Ship of Theseus: Curious Cases of Paraphrasing in LLM-Generated Textscited by 23
  12. 2022MEAL: Stable and Active Learning for Few-Shot Promptingcited by 20
  13. 2023Using GPT-4 to Augment Unbalanced Data for Automatic Scoringcited by 20
  14. 2023Designing LLM Chains by Adapting Techniques from Crowdsourcing Workflowscited by 17
  15. 2021Toward Learning Human-aligned Cross-domain Robust Models by Countering Misaligned Featurescited by 14
  16. 2023Prompting a Large Language Model to Generate Diverse Motivational Messages: A Comparison with Human-Written Messagescited by 10

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 9; 2025: 5; 2026: 3.

17 papers.

Sort:

By area:

  1. 2024A Survey on Data Synthesis and Augmentation for Large Language Modelscited by 41
  2. 2024A Comprehensive Survey on Data Augmentationcited by 40
  3. 2024Improving Demonstration Diversity by Human-Free Fusing for Text-to-SQLcited by 10
  4. 2024LLMs vs Established Text Augmentation Techniques for Classification: When do the Benefits Outweight the Costs?cited by 8
  5. 2024Use Random Selection for Now: Investigation of Few-Shot Selection Strategies in LLM-based Text Augmentation for Classificationcited by 7
  6. 2025Diversity-oriented Data Augmentation with Large Language Modelscited by 7
  7. 2025Accidental Vulnerability: Factors in Fine-Tuning that Shift Model Safeguardscited by 5
  8. 2024A Survey on Automatic Credibility Assessment Using Textual Credibility Signals in the Era of Large Language Modelscited by 3
  9. 2025A Rigorous Evaluation of LLM Data Generation Strategies for Low-Resource Languagescited by 3
  10. 2024Fighting Randomness with Randomness: Mitigating Optimisation Instability of Fine-Tuning using Delayed Ensemble and Noisy Interpolationcited by 2
  11. 2024LLM-DER:A Named Entity Recognition Method Based on Large Language Models for Chinese Coal Chemical Domaincited by 2
  12. 2026Better as Generators Than Classifiers: Leveraging LLMs and Synthetic Data for Low-Resource Multilingual Classificationcited by 2
5 more
  1. 2024Exploring LLM-based Data Annotation Strategies for Medical Dialogue Preference Alignmentcited by 1
  2. 2025CoBA: Counterbias Text Augmentation for Mitigating Various Spurious Correlations via Semantic Triplescited by 1
  3. 2025RoSE: Round-robin Synthetic Data Evaluation for Selecting LLM Generators without Human Test Setscited by 1
  4. 2026Want Better Synthetic Data? Steer It: Activation Steering for Low-Resource Language Generationcited by 0
  5. 2026How Strong Is the Evidence for the Artificial Hivemind? Reevaluating Evidence for the Open-Ended Homogeneity of Language Modelscited by 0