mlsanityData to 14 Oct 2026

LMD3: Language Model Data Density Dependence

arXiv 2405.06331 ↗First posted 10 May 2024cs.LG cs.CLcited by 7

Not on the map. Why?

Abstract (arXiv)

We develop a methodology for analyzing language model task performance at the individual example level based on training data density estimation. Experiments with paraphrasing as a controlled intervention on finetuning data demonstrate that increasing the support in the training distribution for specific test queries results in a measurable increase in density, which is also a significant predictor of the performance increase caused by the intervention. Experiments with pretraining data demonstrate that we can explain a significant fraction of the variance in model perplexity via density measurements. We conclude that our framework can provide statistical evidence of the dependence of a target model's predictions on subsets of its training data, and can more generally be used to characterize the support (or lack thereof) in the training data for a given test task.

Key facts. Cited by 7 papers in the map (data through 14 Oct 2026). Builds on 23 of its 29 references that are in the map. First posted on arXiv 10 May 2024.

Builds on

23 of 29 references.

References · 23

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 13,028
  3. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  4. 2020Measuring Massive Multitask Language Understandingcited by 7,963
  5. 2017Billion-scale similarity search with GPUscited by 2,497
  6. 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,411
  7. 2023Pythia: A Suite for Analyzing Large Language Models Across Training and Scalingcited by 1,786
  8. 2017Understanding Black-box Predictions via Influence Functionscited by 1,741
  9. 2022Quantifying Memorization Across Neural Language Modelscited by 763
  10. 2020Aligning AI With Shared Human Valuescited by 723
  11. 2022Large Language Models Struggle to Learn Long-Tail Knowledgecited by 458
  12. 2023Paraphrasing evades detectors of AI-generated text, but retrieval is an effective defensecited by 429
11 more
  1. 2023Improving Text Embeddings with Large Language Modelscited by 303
  2. 2023TRAK: Attributing Model Behavior at Scalecited by 283
  3. 2023Rethinking Benchmark and Contamination for Language Models with Rephrased Samplescited by 195
  4. 2022Datamodels: Predicting Predictions from Training Datacited by 189
  5. 2023On the Reliability of Watermarks for Large Language Modelscited by 189
  6. 2022Scaling Laws and Interpretability of Learning from Repeated Datacited by 152
  7. 2023D4: Improving LLM Pretraining via Document De-Duplication and Diversificationcited by 150
  8. 2024Investigating Data Contamination for Pre-training Language Modelscited by 91
  9. 2022Datasheet for the Pilecited by 43
  10. 2023Data Similarity is Not Enough to Explain Language Model Performancecited by 9
  11. 2021DEANN: Speeding up Kernel-Density Estimation using Approximate Nearest Neighbor Searchcited by 8

Built on by

Cited by year

Citing papers per year: 2024: 4; 2025: 3.

7 papers.

  1. 2024Generalization v.s. Memorization: Tracing Language Models' Capabilities Back to Pretraining Datacited by 85
  2. 2024Unleashing the Power of Data Tsunami: A Comprehensive Survey on Data Assessment and Selection for Instruction Tuning of Language Modelscited by 29
  3. 2024How Much Can We Forget about Data Contamination?cited by 14
  4. 2025Toward Preference-aligned Large Language Models via Residual-based Model Steeringcited by 3
  5. 2024Leveraging Open Knowledge for Advancing Task Expertise in Large Language Modelscited by 2
  6. 2025Interrogating LLM design under a fair learning doctrinecited by 1
  7. 2025FictionalQA: A Dataset for Studying Memorization and Knowledge Acquisitioncited by 0