Exploring Training Data Attribution under Limited Access Constraints

2025-09-15

AI Summary

Practical training data attribution (TDA) remains challenging under commercial black-box model constraints—where full model access is unavailable—and limited computational resources. Method: We propose a lightweight, high-efficiency attribution framework that avoids reliance on complete model parameters or gradients. Instead, it constructs a compact surrogate model and integrates influence function theory with cross-task attribution transfer, enabling attribution score generation without pretraining on target data. Contribution/Results: To our knowledge, this is the first method to robustly produce discriminative attribution scores under weak black-box access—i.e., forward-only inference, limited queries, or API calls only. Experiments across diverse models and tasks demonstrate substantial improvements over existing black-box TDA baselines: attribution accuracy approaches that of white-box methods, while computational overhead is reduced by one to two orders of magnitude—offering an efficient, deployable solution for model interpretability and data provenance in real-world settings.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Adversarial Attacks & Robustness
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Training data attribution (TDA) plays a critical role in understanding the influence of individual training data points on model predictions. Gradient-based TDA methods, popularized by extit{influence function} for their superior performance, have been widely applied in data selection, data cleaning, data economics, and fact tracing. However, in real-world scenarios where commercial models are not publicly accessible and computational resources are limited, existing TDA methods are often constrained by their reliance on full model access and high computational costs. This poses significant challenges to the broader adoption of TDA in practical applications. In this work, we present a systematic study of TDA methods under various access and resource constraints. We investigate the feasibility of performing TDA under varying levels of access constraints by leveraging appropriately designed solutions such as proxy models. Besides, we demonstrate that attribution scores obtained from models without prior training on the target dataset remain informative across a range of tasks, which is useful for scenarios where computational resources are limited. Our findings provide practical guidance for deploying TDA in real-world environments, aiming to improve feasibility and efficiency under limited access.

Problem

• Studying training data attribution under access constraints
• Addressing computational cost limitations in attribution methods
• Exploring proxy models for feasible data influence analysis

Innovation

• Proxy models for limited access TDA
• Attribution scores without target dataset training
• Practical guidance for efficient TDA deployment