LT-Soups: Bridging Head and Tail Classes via Subsampled Model Soups

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Aminbeidokhti, Masih, Roy, Subhankar, Granger, Eric, Ricci, Elisa, Pedersoli, Marco
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909999591063552
author Aminbeidokhti, Masih
Roy, Subhankar
Granger, Eric
Ricci, Elisa
Pedersoli, Marco
author_facet Aminbeidokhti, Masih
Roy, Subhankar
Granger, Eric
Ricci, Elisa
Pedersoli, Marco
contents Real-world datasets typically exhibit long-tailed (LT) distributions, where a few head classes dominate and many tail classes are severely underrepresented. While recent work shows that parameter-efficient fine-tuning (PEFT) methods like LoRA and AdaptFormer preserve tail-class performance on foundation models such as CLIP, we find that they do so at the cost of head-class accuracy. We identify the head-tail ratio, the proportion of head to tail classes, as a crucial but overlooked factor influencing this trade-off. Through controlled experiments on CIFAR100 with varying imbalance ratio ($ρ$) and head-tail ratio ($η$), we show that PEFT excels in tail-heavy scenarios but degrades in more balanced and head-heavy distributions. To overcome these limitations, we propose LT-Soups, a two-stage model soups framework designed to generalize across diverse LT regimes. In the first stage, LT-Soups averages models fine-tuned on balanced subsets to reduce head-class bias; in the second, it fine-tunes only the classifier on the full dataset to restore head-class accuracy. Experiments across six benchmark datasets show that LT-Soups achieves superior trade-offs compared to both PEFT and traditional model soups across a wide range of imbalance regimes.
format Preprint
id arxiv_https___arxiv_org_abs_2511_10683
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LT-Soups: Bridging Head and Tail Classes via Subsampled Model Soups
Aminbeidokhti, Masih
Roy, Subhankar
Granger, Eric
Ricci, Elisa
Pedersoli, Marco
Machine Learning
Computer Vision and Pattern Recognition
Real-world datasets typically exhibit long-tailed (LT) distributions, where a few head classes dominate and many tail classes are severely underrepresented. While recent work shows that parameter-efficient fine-tuning (PEFT) methods like LoRA and AdaptFormer preserve tail-class performance on foundation models such as CLIP, we find that they do so at the cost of head-class accuracy. We identify the head-tail ratio, the proportion of head to tail classes, as a crucial but overlooked factor influencing this trade-off. Through controlled experiments on CIFAR100 with varying imbalance ratio ($ρ$) and head-tail ratio ($η$), we show that PEFT excels in tail-heavy scenarios but degrades in more balanced and head-heavy distributions. To overcome these limitations, we propose LT-Soups, a two-stage model soups framework designed to generalize across diverse LT regimes. In the first stage, LT-Soups averages models fine-tuned on balanced subsets to reduce head-class bias; in the second, it fine-tunes only the classifier on the full dataset to restore head-class accuracy. Experiments across six benchmark datasets show that LT-Soups achieves superior trade-offs compared to both PEFT and traditional model soups across a wide range of imbalance regimes.
title LT-Soups: Bridging Head and Tail Classes via Subsampled Model Soups
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.10683