STRIDE: Training Data Attribution via Sparse Recovery from Subset Perturbations

Hugging Face Daily Papers Papers

Summary

STRIDE is a new framework for training data attribution in LLMs that models functional effects in activation space using sparse recovery and steering operators, achieving state-of-the-art accuracy with 13x speedup over previous methods.

Training Data Attribution (TDA) seeks to trace a model's predictions back to its training data. The gold standard for TDA relies on causal interventions, observing how a model changes when data is added or removed, but repeated retraining is computationally challenging for Large Language Models (LLMs). Consequently, most approaches approximate this effect in the parameter space using gradients. However, tracking gradients across billions of parameters is not only prohibitively expensive but relies on local approximations. In this work, we propose a shift: rather than estimating parameter changes, we model the functional effect of training data in the activation space. We introduce STRIDE (Steering-based Training Data Influence Decomposition), a framework that formulates TDA as a sparse recovery problem in the spirit of compressive sensing. STRIDE learns lightweight "steering operators" that mimic the behavioral shift caused by training on data subsets. By measuring how these operators perturb test predictions, we recover individual training example influences via sparse linear decomposition. STRIDE achieves state-of-the-art for LLM pre-training attribution while being an order of magnitude (13times) faster than previous art. We further validate its practical utility through downstream applications including data selection, data contamination, and qualitative analysis.
Original Article
View Cached Full Text

Cached at: 06/04/26, 03:41 AM

Paper page - STRIDE: Training Data Attribution via Sparse Recovery from Subset Perturbations

Source: https://huggingface.co/papers/2606.05165

Abstract

STRIDE framework enables efficient training data attribution for LLMs by modeling functional effects in activation space through sparse recovery and steering operators, achieving superior speed and accuracy compared to traditional gradient-based methods.

Training Data Attribution(TDA) seeks to trace a model’s predictions back to its training data. The gold standard for TDA relies oncausal interventions, observing how a model changes when data is added or removed, but repeated retraining is computationally challenging forLarge Language Models(LLMs). Consequently, most approaches approximate this effect in theparameter spaceusinggradients. However, trackinggradientsacross billions of parameters is not only prohibitively expensive but relies on local approximations. In this work, we propose a shift: rather than estimating parameter changes, we model the functional effect of training data in theactivation space. We introduce STRIDE (Steering-based Training Data Influence Decomposition), a framework that formulates TDA as asparse recoveryproblem in the spirit ofcompressive sensing. STRIDE learns lightweight “steering operators” that mimic the behavioral shift caused by training on data subsets. By measuring how these operators perturb test predictions, we recover individual training example influences viasparse linear decomposition. STRIDE achieves state-of-the-art for LLM pre-training attribution while being an order of magnitude (13times) faster than previous art. We further validate its practical utility through downstream applications including data selection, data contamination, and qualitative analysis.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2606\.05165

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2606.05165 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2606.05165 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2606.05165 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

DataDignity: Training Data Attribution for Large Language Models

arXiv cs.AI

This paper introduces DataDignity, a framework and benchmark (FakeWiki) for pinpoint provenance, aiming to identify the specific training data sources that support an LLM's response. It proposes ScoringModel and SteerFuse methods to improve attribution accuracy over standard retrieval baselines.