TRAK: Attributing Model Behavior at Scale
This work addresses the challenge of tracing model predictions back to training data for researchers and practitioners using large-scale, differentiable models, offering a practical solution that balances efficacy and computational cost.
The authors tackled the problem of data attribution for large-scale models, where existing methods are either computationally tractable but ineffective or effective but impractical. They introduced TRAK, which matches the performance of methods requiring thousands of models while using only a handful, demonstrating its utility on models like ImageNet classifiers, CLIP, BERT, and mT5.
The goal of data attribution is to trace model predictions back to training data. Despite a long line of work towards this goal, existing approaches to data attribution tend to force users to choose between computational tractability and efficacy. That is, computationally tractable methods can struggle with accurately attributing model predictions in non-convex settings (e.g., in the context of deep neural networks), while methods that are effective in such regimes require training thousands of models, which makes them impractical for large models or datasets. In this work, we introduce TRAK (Tracing with the Randomly-projected After Kernel), a data attribution method that is both effective and computationally tractable for large-scale, differentiable models. In particular, by leveraging only a handful of trained models, TRAK can match the performance of attribution methods that require training thousands of models. We demonstrate the utility of TRAK across various modalities and scales: image classifiers trained on ImageNet, vision-language models (CLIP), and language models (BERT and mT5). We provide code for using TRAK (and reproducing our work) at https://github.com/MadryLab/trak .