mlsanityData to 14 Oct 2026

f-INE: A Hypothesis Testing Framework for Estimating Influence under Training Randomness

arXiv 2510.10510 ↗First posted 12 Oct 2025cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Influence estimation methods promise to explain and debug machine learning by estimating the impact of individual samples on the final model. Yet, existing methods collapse under training randomness: the same example may appear critical in one run and irrelevant in the next. Such instability undermines their use in data curation or cleanup since it is unclear if we indeed deleted/kept the correct datapoints. To overcome this, we introduce *f-influence* -- a new influence estimation framework grounded in hypothesis testing that explicitly accounts for training randomness, and establish desirable properties that make it suitable for reliable influence estimation. We also design a highly efficient algorithm **f**-**IN**fluence **E**stimation (**f-INE**) that computes f-influence **in a single training run**. Finally, we scale up f-INE to estimate influence of instruction tuning data on Llama-3.1-8B and show it can reliably detect poisoned samples that steer model opinions, demonstrating its utility for data cleanup and attributing model behavior.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 39 of its 55 references that are in the map. First posted on arXiv 12 Oct 2025.

Builds on

39 of 55 references.

References · 39

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  3. 2016"Why Should I Trust You?": Explaining the Predictions of Any Classifiercited by 5,070
  4. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
  5. 2016Membership Inference Attacks against Machine Learning Modelscited by 2,281
  6. 2017Understanding Black-box Predictions via Influence Functionscited by 1,741
  7. 2021Membership Inference Attacks From First Principlescited by 878
  8. 2019Data Shapley: Equitable Valuation of Data for Machine Learningcited by 629
  9. 2020Estimating Training Data Influence by Tracing Gradient Descentcited by 580
  10. 2024LESS: Selecting Influential Data for Targeted Instruction Tuningcited by 532
  11. 2020What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimationcited by 497
  12. 2023TRAK: Attributing Model Behavior at Scalecited by 283
27 more
  1. 2019Towards Efficient Data Valuation Based on the Shapley Valuecited by 281
  2. 2020Influence Functions in Deep Learning Are Fragilecited by 252
  3. 2016What makes ImageNet good for transfer learning?cited by 192
  4. 2023Backdooring Instruction-Tuned Large Language Models with Virtual Prompt Injectioncited by 157
  5. 2019Efficient Task-Specific Data Valuation for Nearest Neighbor Algorithmscited by 148
  6. 2022If Influence Functions are the Answer, Then What is the Question?cited by 144
  7. 2022Training Data Influence Analysis and Estimation: A Surveycited by 135
  8. 2021Scaling Up Influence Functionscited by 132
  9. 2022Data Banzhaf: A Robust Data Valuation Framework for Machine Learningcited by 132
  10. 2023Privacy Auditing with One (1) Training Runcited by 123
  11. 2024Data Shapley in One Training Runcited by 81
  12. 2020A Distributional Framework for Data Valuationcited by 80
  13. 2023LAVA: Data Valuation without Pre-Specified Learning Algorithmscited by 69
  14. 2013Bounding the Estimation Error of Sampling-based Shapley Value Approximationcited by 58
  15. 2016Auditing Black-box Models for Indirect Influencecited by 41
  16. 2021Revisiting Methods for Finding Influential Examplescited by 39
  17. 2020Efficient computation and analysis of distributional Shapley valuescited by 38
  18. 2023Data-OOB: Out-of-bag Estimate as a Simple and Efficient Data Valuecited by 38
  19. 2022CS-Shapley: Class-wise Shapley Values for Data Valuation in Classificationcited by 32
  20. 2023Theoretical and Practical Perspectives on what Influence Functions Docited by 29
  21. 2024Rethinking Data Shapley for Data Selection Tasks: Misleads and Meritscited by 29
  22. 2021Rethinking Influence Functions of Neural Networks in the Over-parameterized Regimecited by 26
  23. 2025Privacy Auditing of Large Language Modelscited by 23
  24. 2023On the Variance of Neural Network Training with respect to Test Sets and Distributionscited by 22
  25. 2022Variance reduced Shapley value estimation for trustworthy data valuationcited by 16
  26. 2023Low-rank Adaptation of Large Language Model Rescoring for Parameter-Efficient Speech Recognitioncited by 13
  27. 2025The Surprising Effectiveness of Membership Inference with Simple N-Gram Coveragecited by 7

Built on by

Not cited yet.