Metraipolis › archive › dc_698ecd4dcdeb

MoR-MLLM: Mixture of Recursions for Efficient Multimodal Large Language Models — arXiv:2610.08830

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.08830
License (as archived): CC0-1.0
Archived: 2026-10-08T12:46:49.228Z
SHA-256 of the archived text: aeb36c732b8eef0a2ddb5609e4cdd6e89fd90ceb3defaf013fe9acffe09cbe45
Stable URL of this page: https://metraipolis.com/documento/dc_698ecd4dcdeb

How to cite

MoR-MLLM: Mixture of Recursions for Efficient Multimodal Large Language Models — arXiv:2610.08830 — arXiv (en), https://arxiv.org/abs/2610.08830 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_698ecd4dcdeb (Metraipolis archive dc_698ecd4dcdeb, sha256 aeb36c732b8eef0a2ddb5609e4cdd6e89fd90ceb3defaf013fe9acffe09cbe45)

BibTeX

@misc{dc698ecd4dcdeb,
  title = {MoR-MLLM: Mixture of Recursions for Efficient Multimodal Large Language Models — arXiv:2610.08830},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.08830},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_698ecd4dcdeb; sha256 aeb36c732b8eef0a2ddb5609e4cdd6e89fd90ceb3defaf013fe9acffe09cbe45},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.08830. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Multimodal Large Language Models (MLLMs) have demonstrated remarkable reasoning capabilities across vision and language tasks. However, their massive computational and memory demands hinder real-world deployment. While recent efforts reduce costs by employing lightweight language backbones, existing paradigms remain computation-dense due to their static sparsity and depth allocation, which cannot adapt to the semantic complexity of each token. To this end, we propose MoR-MLLM, a computation-sparse MLLM based on the recent Mixture-of-Recursions (MoR) framework. MoR-MLLM introduces adaptive per-token recursion, allowing the model to dynamically adjust its recursive depth and allocate more computation to visually or linguistically challenging tokens while skipping redundant operations for simpler ones. To stabilize the training of recursive sparsity in multimodal settings, we further design a three-stage MoR-Tuning strategy and an entropy-regularized loss to encourage diverse routing distributions. Extensive experiments show that compared with recent advanced tiny MLLMs, our proposed MoR-MLLM can greatly reduce the training memory and computation complexity while retaining high performance on various vision-language tasks.

Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2610.08830 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-08T12:18:15.271Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).