English

GraSS: Scalable Data Attribution with Gradient Sparsification and Sparse Projection

Machine Learning 2025-10-29 v3 Artificial Intelligence

Abstract

Gradient-based data attribution methods, such as influence functions, are critical for understanding the impact of individual training samples without requiring repeated model retraining. However, their scalability is often limited by the high computational and memory costs associated with per-sample gradient computation. In this work, we propose GraSS, a novel gradient compression algorithm and its variants FactGraSS for linear layers specifically, that explicitly leverage the inherent sparsity of per-sample gradients to achieve sub-linear space and time complexity. Extensive experiments demonstrate the effectiveness of our approach, achieving substantial speedups while preserving data influence fidelity. In particular, FactGraSS achieves up to 165% faster throughput on billion-scale models compared to the previous state-of-the-art baselines. Our code is publicly available at https://github.com/TRAIS-Lab/GraSS.

Keywords

Cite

@article{arxiv.2505.18976,
  title  = {GraSS: Scalable Data Attribution with Gradient Sparsification and Sparse Projection},
  author = {Pingbang Hu and Joseph Melkonian and Weijing Tang and Han Zhao and Jiaqi W. Ma},
  journal= {arXiv preprint arXiv:2505.18976},
  year   = {2025}
}

Comments

Accepted at the 39th Conference on Neural Information Processing Systems (NeurIPS 2025)

R2 v1 2026-07-01T02:36:46.147Z