Saved in:
Bibliographic Details
Main Authors: Ravé, Simon, Rasti, Pejman, Rousseau, David
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.05718
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918431087919104
author Ravé, Simon
Rasti, Pejman
Rousseau, David
author_facet Ravé, Simon
Rasti, Pejman
Rousseau, David
contents Decreasing sequence length is a common way to accelerate transformers, but prior token reduction work often targets classification and reports proxy metrics rather than end-to-end latency. For semantic segmentation, token reduction is further constrained by the need to reconstruct dense, pixel-aligned features, and on modern accelerators the overhead of computing merge maps can erase expected gains. We propose Mutual Pair Merging (MPM), a training-free token aggregation module that forms mutual nearest-neighbor pairs in cosine space, averages each pair, and records a merge map enabling a gather-based reconstruction before the decoder so that existing segmentation heads can be used unchanged. MPM introduces no learned parameters and no continuous compression knob (no keep-rate or threshold). The speed-accuracy trade-off is set by a discrete insertion schedule. We benchmark end-to-end latency on an NVIDIA H100 GPU (with and without FlashAttention-2) and a Raspberry Pi 5 across standard segmentation datasets. On ADE20K, MPM reduces per-image latency by up to 60% for ViT-Tiny on Raspberry Pi 5, and increases throughput by up to 20% on H100 with FlashAttention-2 while keeping the mIoU drop below 3%. These results suggest that simple, reconstruction-aware, training-free token merging can translate into practical wall-clock gains for segmentation when overhead is explicitly accounted for.
format Preprint
id arxiv_https___arxiv_org_abs_2604_05718
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MPM: Mutual Pair Merging for Efficient Vision Transformers
Ravé, Simon
Rasti, Pejman
Rousseau, David
Computer Vision and Pattern Recognition
Decreasing sequence length is a common way to accelerate transformers, but prior token reduction work often targets classification and reports proxy metrics rather than end-to-end latency. For semantic segmentation, token reduction is further constrained by the need to reconstruct dense, pixel-aligned features, and on modern accelerators the overhead of computing merge maps can erase expected gains. We propose Mutual Pair Merging (MPM), a training-free token aggregation module that forms mutual nearest-neighbor pairs in cosine space, averages each pair, and records a merge map enabling a gather-based reconstruction before the decoder so that existing segmentation heads can be used unchanged. MPM introduces no learned parameters and no continuous compression knob (no keep-rate or threshold). The speed-accuracy trade-off is set by a discrete insertion schedule. We benchmark end-to-end latency on an NVIDIA H100 GPU (with and without FlashAttention-2) and a Raspberry Pi 5 across standard segmentation datasets. On ADE20K, MPM reduces per-image latency by up to 60% for ViT-Tiny on Raspberry Pi 5, and increases throughput by up to 20% on H100 with FlashAttention-2 while keeping the mIoU drop below 3%. These results suggest that simple, reconstruction-aware, training-free token merging can translate into practical wall-clock gains for segmentation when overhead is explicitly accounted for.
title MPM: Mutual Pair Merging for Efficient Vision Transformers
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.05718