Scaling Laws for Native Multimodal Models

• URL: http://arxiv.org/abs/2504.07951v2
• Date: Fri, 11 Apr 2025 06:35:42 GMT
• Title: Scaling Laws for Native Multimodal Models
• Authors: Mustafa Shukor, Enrico Fini, Victor Guilherme Turrisi da Costa, Matthieu Cord, Joshua Susskind, Alaaeldin El-Nouby,
• Abstract summary: We revisit the architectural design of native multimodal models and conduct an extensive scaling laws study.<n>Our investigation reveals no inherent advantage to late-fusion architectures over early-fusion ones.<n>We show that incorporating Mixture of Experts (MoEs) allows for models that learn modality-specific weights, significantly enhancing performance.
• Score: 53.490942903659565
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Building general-purpose models that can effectively perceive the world through multimodal signals has been a long-standing goal. Current approaches involve integrating separately pre-trained components, such as connecting vision encoders to LLMs and continuing multimodal training. While such approaches exhibit remarkable sample efficiency, it remains an open question whether such late-fusion architectures are inherently superior. In this work, we revisit the architectural design of native multimodal models (NMMs)--those trained from the ground up on all modalities--and conduct an extensive scaling laws study, spanning 457 trained models with different architectures and training mixtures. Our investigation reveals no inherent advantage to late-fusion architectures over early-fusion ones, which do not rely on image encoders. On the contrary, early-fusion exhibits stronger performance at lower parameter counts, is more efficient to train, and is easier to deploy. Motivated by the strong performance of the early-fusion architectures, we show that incorporating Mixture of Experts (MoEs) allows for models that learn modality-specific weights, significantly enhancing performance.

Related papers

• An Efficient and Mixed Heterogeneous Model for Image Restoration [71.85124734060665]
  Current mainstream approaches are based on three architectural paradigms: CNNs, Transformers, and Mambas. We propose RestorMixer, an efficient and general-purpose IR model based on mixed-architecture fusion.
  arXiv  Detail & Related papers  (2025-04-15T08:19:12Z)
• MIND: Modality-Informed Knowledge Distillation Framework for Multimodal Clinical Prediction Tasks [50.98856172702256]
  We propose the Modality-INformed knowledge Distillation (MIND) framework, a multimodal model compression approach.<n>MIND transfers knowledge from ensembles of pre-trained deep neural networks of varying sizes into a smaller multimodal student.<n>We evaluate MIND on binary and multilabel clinical prediction tasks using time series data and chest X-ray images.
  arXiv  Detail & Related papers  (2025-02-03T08:50:00Z)
• Uni-MoE: Scaling Unified Multimodal LLMs with Mixture of Experts [54.529880848937104]
  We develop a unified MLLM with the MoE architecture, named Uni-MoE, that can handle a wide array of modalities. Specifically, it features modality-specific encoders with connectors for a unified multimodal representation. We evaluate the instruction-tuned Uni-MoE on a comprehensive set of multimodal datasets.
  arXiv  Detail & Related papers  (2024-05-18T12:16:01Z)
• Improving Efficiency of Diffusion Models via Multi-Stage Framework and Tailored Multi-Decoder Architectures [12.703947839247693]
  Diffusion models, emerging as powerful deep generative tools, excel in various applications. However, their remarkable generative performance is hindered by slow training and sampling. This is due to the necessity of tracking extensive forward and reverse diffusion trajectories. We present a multi-stage framework inspired by our empirical findings to tackle these challenges.
  arXiv  Detail & Related papers  (2023-12-14T17:48:09Z)
• Improving Discriminative Multi-Modal Learning with Large-Scale Pre-Trained Models [51.5543321122664]
  This paper investigates how to better leverage large-scale pre-trained uni-modal models to enhance discriminative multi-modal learning. We introduce Multi-Modal Low-Rank Adaptation learning (MMLoRA)
  arXiv  Detail & Related papers  (2023-10-08T15:01:54Z)
• Multi-Architecture Multi-Expert Diffusion Models [18.463425624382115]
  We introduce Multi-architecturE Multi-Expert diffusion models (MEME) MEME operates 3.3 times faster than baselines while improving image generation quality (FID scores) by 0.62 (FFHQ) and 0.37 (CelebA) We argue that MEME opens a new design choice for diffusion models that can be easily applied in other scenarios, such as large multi-expert models.
  arXiv  Detail & Related papers  (2023-06-08T07:24:08Z)
• An Empirical Study of Multimodal Model Merging [148.48412442848795]
  Model merging is a technique that fuses multiple models trained on different tasks to generate a multi-task solution. We conduct our study for a novel goal where we can merge vision, language, and cross-modal transformers of a modality-specific architecture. We propose two metrics that assess the distance between weights to be merged and can serve as an indicator of the merging outcomes.
  arXiv  Detail & Related papers  (2023-04-28T15:43:21Z)
• eP-ALM: Efficient Perceptual Augmentation of Language Models [70.47962271121389]
  We propose to direct effort to efficient adaptations of existing models, and propose to augment Language Models with perception. Existing approaches for adapting pretrained models for vision-language tasks still rely on several key components that hinder their efficiency. We show that by freezing more than 99% of total parameters, training only one linear projection layer, and prepending only one trainable token, our approach (dubbed eP-ALM) significantly outperforms other baselines on VQA and Captioning.
  arXiv  Detail & Related papers  (2023-03-20T19:20:34Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.