Interaction-Aware Influence Functions for Group Attribution

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Heo, Jaeseung, Yun, Kyeongheung, Choi, Youngbin, Hwang, Sehyun, Ok, Jungseul, Kim, Dongwoo
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916015723511808
author Heo, Jaeseung
Yun, Kyeongheung
Choi, Youngbin
Hwang, Sehyun
Ok, Jungseul
Kim, Dongwoo
author_facet Heo, Jaeseung
Yun, Kyeongheung
Choi, Youngbin
Hwang, Sehyun
Ok, Jungseul
Kim, Dongwoo
contents Influence functions approximate how removing a training example changes a quantity of interest, called the target function, such as a held-out loss. To estimate the influence of a group of examples, the standard practice is to sum the individual influences of its members. However, this sum does not capture how examples jointly affect the target: a pair of examples may be redundant or complementary, but the sum cannot distinguish these cases. We propose an interaction-aware influence function that characterizes how interactions between examples influence the target. By expanding the target to second order around the trained parameters, we obtain an estimator that augments the standard sum with a pairwise interaction term that captures the alignment between two examples' effects on the target. We empirically evaluate our estimator in two settings. First, on six dataset-model pairs spanning logistic regression, MLPs, and ResNet-9, our estimator tracks leave-group-out retraining substantially better than first-order influence across all settings. Second, when used as a greedy selection rule for instruction-tuning data on Llama-3.1-8B, it beats prior influence-based and representation-similarity baselines on five of seven downstream tasks, in a regime where standard influence-based selection underperforms random selection.
format Preprint
id arxiv_https___arxiv_org_abs_2605_15675
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Interaction-Aware Influence Functions for Group Attribution
Heo, Jaeseung
Yun, Kyeongheung
Choi, Youngbin
Hwang, Sehyun
Ok, Jungseul
Kim, Dongwoo
Machine Learning
Artificial Intelligence
Influence functions approximate how removing a training example changes a quantity of interest, called the target function, such as a held-out loss. To estimate the influence of a group of examples, the standard practice is to sum the individual influences of its members. However, this sum does not capture how examples jointly affect the target: a pair of examples may be redundant or complementary, but the sum cannot distinguish these cases. We propose an interaction-aware influence function that characterizes how interactions between examples influence the target. By expanding the target to second order around the trained parameters, we obtain an estimator that augments the standard sum with a pairwise interaction term that captures the alignment between two examples' effects on the target. We empirically evaluate our estimator in two settings. First, on six dataset-model pairs spanning logistic regression, MLPs, and ResNet-9, our estimator tracks leave-group-out retraining substantially better than first-order influence across all settings. Second, when used as a greedy selection rule for instruction-tuning data on Llama-3.1-8B, it beats prior influence-based and representation-similarity baselines on five of seven downstream tasks, in a regime where standard influence-based selection underperforms random selection.
title Interaction-Aware Influence Functions for Group Attribution
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2605.15675