mlsanityData to 7 Oct 2026

Evading Data Contamination Detection for Language Models is (too) Easy

arXiv 2402.02823 ↗First posted 5 Feb 2024cs.LG cs.AI cs.CLcited by 30

Not on the map. Why?

Abstract (arXiv)

Large language models are widespread, with their performance on benchmarks frequently guiding user preferences for one model over another. However, the vast amount of data these models are trained on can inadvertently lead to contamination with public benchmarks, thus compromising performance measurements. While recently developed contamination detection methods try to address this issue, they overlook the possibility of deliberate contamination by malicious model providers aiming to evade detection. We argue that this setting is of crucial importance as it casts doubt on the reliability of public benchmarks. To more rigorously study this issue, we propose a categorization of both model providers and contamination detection methods. This reveals vulnerabilities in existing methods that we exploit with EAL, a simple yet effective contamination technique that significantly inflates benchmark performance while completely evading current detection methods.

Key facts. Cited by 30 papers in the map (data through 7 Oct 2026). Builds on 37 of its 52 references that are in the map. First posted on arXiv 5 Feb 2024.

Builds on

37 of 52 references.

References · 37

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2023GPT-4 Technical Reportcited by 18,132
  3. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  4. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  5. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  6. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  7. 2020Measuring Massive Multitask Language Understandingcited by 7,768
  8. 2018Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challengecited by 4,710
  9. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,919
  10. 2020The Pile: An 800GB Dataset of Diverse Text for Language Modelingcited by 2,346
  11. 2020Extracting Training Data from Large Language Modelscited by 2,144
  12. 2023A Survey on Evaluation of Large Language Modelscited by 1,227
25 more
  1. 2021Membership Inference Attacks From First Principlescited by 839
  2. 2023The RefinedWeb Dataset for Falcon LLM: Outperforming Curated Corpora with Web Data, and Web Data Onlycited by 755
  3. 2021Deduplicating Training Data Makes Language Models Bettercited by 705
  4. 2021Documenting Large Webtext Corpora: A Case Study on the Colossal Clean Crawled Corpuscited by 481
  5. 2023Detecting Pretraining Data from Large Language Modelscited by 375
  6. 2023NLP Evaluation in trouble: On the Need to Measure LLM Data Contamination for each Benchmarkcited by 339
  7. 2021Experts, Errors, and Context: A Large-Scale Study of Human Evaluation for Machine Translationcited by 333
  8. 2023Membership Inference Attacks against Language Models via Neighbourhood Comparisoncited by 249
  9. 2023Proving Test Set Contamination in Black Box Language Modelscited by 227
  10. 2023Rethinking Benchmark and Contamination for Language Models with Rephrased Samplescited by 192
  11. 2023Time Travel in LLMs: Tracing Data Contamination in Large Language Modelscited by 175
  12. 2022Quantifying Privacy Risks of Masked Language Models Using Membership Inference Attackscited by 168
  13. 2023Platypus: Quick, Cheap, and Powerful Refinement of LLMscited by 150
  14. 2021On the Importance of Difficulty Calibration in Membership Inference Attackscited by 137
  15. 2023Investigating Data Contamination in Modern Benchmarks for Large Language Modelscited by 136
  16. 2023Task Contamination: Language Models May Not Be Few-Shot Anymorecited by 98
  17. 2017Paraphrase Generation with Deep Reinforcement Learningcited by 58
  18. 2023LatestEval: Addressing Data Contamination in Language Model Evaluation through Dynamic and Time-Sensitive Test Constructioncited by 57
  19. 2023Estimating Contamination via Perplexity: Quantifying Memorisation in Language Model Evaluationcited by 47
  20. 2023An Open Source Data Contamination Report for Large Language Modelscited by 47
  21. 2023Data Contamination Quiz: A Tool to Detect and Estimate Contamination in Large Language Modelscited by 46
  22. 2023Data Contamination Through the Lens of Timecited by 39
  23. 2023Competition-Level Problems are Effective LLM Evaluatorscited by 25
  24. 2023Koala: An Index for Quantifying Overlaps with Pre-training Corporacited by 19
  25. 2023CLEAN-EVAL: Clean Evaluation on Contaminated Large Language Modelscited by 13

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 12; 2025: 11; 2026: 7.

30 papers.

Sort:

By area:

  1. 2024ConStat: Performance-Based Contamination Detection in Large Language Modelscited by 33
  2. 2025A Survey on Data Contamination for Large Language Modelscited by 33
  3. 2024Unveiling the Spectrum of Data Contamination in Language Models: A Survey from Detection to Remediationcited by 30
  4. 2024VarBench: Robust Language Model Benchmarking Through Dynamic Variable Perturbationcited by 29
  5. 2024Does Data Contamination Detection Work (Well) for LLMs? A Survey and Evaluation on Detection Assumptionscited by 26
  6. 2025Language Models May Verbatim Complete Text They Were Not Explicitly Trained Oncited by 20
  7. 2024TabEBM: A Tabular Data Augmentation Method with Distinct Class-Specific Energy-Based Modelscited by 15
  8. 2024Ward: Provable RAG Dataset Inference via LLM Watermarkscited by 15
  9. 2025LastingBench: Defend Benchmarks Against Knowledge Leakagecited by 14
  10. 2025The Emperor's New Clothes in Benchmarking? A Rigorous Examination of Mitigation Strategies for LLM Benchmark Data Contaminationcited by 13
  11. 2024A Comprehensive Survey of Contamination Detection Methods in Large Language Modelscited by 12
  12. 2024LLM-as-an-Interviewer: Beyond Static Testing Through Dynamic LLM Evaluationcited by 12
18 more
  1. 2024DICE: Detecting In-distribution Contamination in LLM's Fine-tuning Phase for Math Reasoningcited by 11
  2. 2025On The Fragility of Benchmark Contamination Detection in Reasoning Modelscited by 8
  3. 2025Where Did This Sentence Come From? Tracing Provenance in LLM Reasoning Distillationcited by 6
  4. 2024LSHBloom: Memory-efficient, Extreme-scale Document Deduplicationcited by 5
  5. 2024Polyrating: A Cost-Effective and Bias-Aware Rating System for LLM Evaluationcited by 3
  6. 2025Establishing Trustworthy LLM Evaluation via Shortcut Neuron Analysiscited by 3
  7. 2025CoreEval: Automatically Building Contamination-Resilient Datasets with Real-World Knowledge toward Reliable LLM Evaluationcited by 3
  8. 2025LLMEval-Fair: A Large-Scale Longitudinal Study on Robust and Fair Evaluation of Large Language Modelscited by 2
  9. 2026The Illusion of Reasoning: Exposing Evasive Data Contamination in LLMs via Zero-CoT Truncationcited by 2
  10. 2024CAP: Data Contamination Detection via Consistency Amplificationcited by 1
  11. 2025AdEval: Alignment-based Dynamic Evaluation to Mitigate Data Contamination in Large Language Modelscited by 1
  12. 2025ArenaBencher: Automatic Benchmark Evolution via Multi-Model Competitive Evaluationcited by 1
  13. 2026When Agents Look the Same: Quantifying Distillation-Induced Similarity in Tool-Use Behaviorscited by 1
  14. 2026TRACER: A Semantic-Aware Framework for Fine-Grained Contamination Detection in Code LLMscited by 1
  15. 2026Pretraining Data Exposure in Large Language Models: A Survey of Membership Inference, Data Contamination, and Security Implicationscited by 1
  16. 2026Learning to Detect Language Model Training Data via Active Reconstructioncited by 0
  17. 2026When Is Benchmark Contamination Detectable? Information Limits and Power-Calibrated Auditscited by 0
  18. 2026CleanScore: Black-Box Benchmark Audits with Negative Controls and Sensitivity Boundscited by 0