Looping Back to Move Forward: Recursive Transformers for Efficient and Flexible Large Multimodal Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Xu, Ruihan, Gao, Yuting, Wang, Lan, Li, Jianing, Chen, Weihao, Guo, Qingpei, Yang, Ming, Zhang, Shiliang
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918328926208000
author Xu, Ruihan
Gao, Yuting
Wang, Lan
Li, Jianing
Chen, Weihao
Guo, Qingpei
Yang, Ming
Zhang, Shiliang
author_facet Xu, Ruihan
Gao, Yuting
Wang, Lan
Li, Jianing
Chen, Weihao
Guo, Qingpei
Yang, Ming
Zhang, Shiliang
contents Large Multimodal Models (LMMs) have achieved remarkable success in vision-language tasks, yet their vast parameter counts are often underutilized during both training and inference. In this work, we embrace the idea of looping back to move forward: reusing model parameters through recursive refinement to extract stronger multimodal representations without increasing model size. We propose RecursiveVLM, a recursive Transformer architecture tailored for LMMs. Two key innovations enable effective looping: (i) a Recursive Connector that aligns features across recursion steps by fusing intermediate-layer hidden states and applying modality-specific projections, respecting the distinct statistical structures of vision and language tokens; (ii) a Monotonic Recursion Loss that supervises every step and guarantees performance improves monotonically with recursion depth. This design transforms recursion into an on-demand refinement mechanism: delivering strong results with few loops on resource-constrained devices and progressively improving outputs when more computation resources are available. Experiments show consistent gains of +3% over standard Transformers and +7% over vanilla recursive baselines, demonstrating that strategic looping is a powerful path toward efficient, deployment-adaptive LMMs.
format Preprint
id arxiv_https___arxiv_org_abs_2602_09080
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Looping Back to Move Forward: Recursive Transformers for Efficient and Flexible Large Multimodal Models
Xu, Ruihan
Gao, Yuting
Wang, Lan
Li, Jianing
Chen, Weihao
Guo, Qingpei
Yang, Ming
Zhang, Shiliang
Machine Learning
Artificial Intelligence
Large Multimodal Models (LMMs) have achieved remarkable success in vision-language tasks, yet their vast parameter counts are often underutilized during both training and inference. In this work, we embrace the idea of looping back to move forward: reusing model parameters through recursive refinement to extract stronger multimodal representations without increasing model size. We propose RecursiveVLM, a recursive Transformer architecture tailored for LMMs. Two key innovations enable effective looping: (i) a Recursive Connector that aligns features across recursion steps by fusing intermediate-layer hidden states and applying modality-specific projections, respecting the distinct statistical structures of vision and language tokens; (ii) a Monotonic Recursion Loss that supervises every step and guarantees performance improves monotonically with recursion depth. This design transforms recursion into an on-demand refinement mechanism: delivering strong results with few loops on resource-constrained devices and progressively improving outputs when more computation resources are available. Experiments show consistent gains of +3% over standard Transformers and +7% over vanilla recursive baselines, demonstrating that strategic looping is a powerful path toward efficient, deployment-adaptive LMMs.
title Looping Back to Move Forward: Recursive Transformers for Efficient and Flexible Large Multimodal Models
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2602.09080