Rethinking UMM Visual Generation: Masked Modeling for Efficient Image-Only Pre-training

Dimensions overall score 9.0

• The proposed Image-Only Training for UMMs (IOMM) framework improves training efficiency by using abundant unlabeled image-only data for pre-training, removing dependency on paired data for this phase.method 95%
  Evidencepartial

  The first stage pre-trains the visual generative component exclusively using abundant unlabeled image-only data, thereby removing the dependency on paired data for this costly phase.

  Implicationmissing

  Implication not extracted yet.

  Verificationpartial

  partial

• IOMM-B (3.6B) model achieves state-of-the-art performance with 0.89 on GenEval and 0.55 on WISE, surpassing BAGEL-7B (0.82 & 0.55) and BLIP3-o-4B (0.84 & 0.50).result 95%
  Evidencepartial

  It achieves 0.89 on GenEval and 0.55 on WISE--surpassing strong baselines such as BAGEL-7B (0.82 & 0.55) and BLIP3-o-4B (0.84 & 0.50).

  Implicationmissing

  Implication not extracted yet.

  Verificationpartial

  partial

• IOMM-B (3.6B) model was trained from scratch using only approximately 1050 H800 GPU hours, with 1000 hours dedicated to the efficient image-only pre-training stage.technical 95%
  Evidencepartial

  For example, our IOMM-B (3.6B) model was trained from scratch using only ~1050 H800 GPU hours (with the vast majority, 1000 hours, dedicated to the efficient image-only pre-training stage).

  Implicationmissing

  Implication not extracted yet.

  Verificationpartial

  partial

• The second stage fine-tunes the model using a mixture of unlabeled images and a small curated set of text-image pairs, leading to improved instruction alignment and generative quality.method 90%
  Evidencepartial

  The second stage fine-tunes the model using a mixture of unlabeled images and a small curated set of text-image pairs, leading to improved instruction alignment and generative quality.

  Implicationmissing

  Implication not extracted yet.

  Verificationpartial

  partial

• Unified Multimodal Models (UMMs) are constrained by inefficient pre-training paradigms and scarce, high-quality text-image paired data for their visual generation components.limitation 90%
  Evidencepartial

  Unified Multimodal Models (UMMs) are often constrained by the pre-training of their visual generation components, which typically relies on inefficient paradigms and scarce, high-quality text-image paired data.

  Implicationmissing

  Implication not extracted yet.

  Verificationpartial

  partial

• The IOMM framework addresses bottlenecks in UMM visual generation pre-training by using a data-efficient two-stage training approach.technical 90%
  Evidencepartial

  To address them, we propose Image-Only Training for UMMs (IOMM), a data-efficient two-stage training framework.

  Implicationmissing

  Implication not extracted yet.

  Verificationpartial

  partial

• Extensive experiments show that IOMM not only improves training efficiency but also achieves state-of-the-art (SOTA) performance.result 90%
  Evidencepartial

  Extensive experiments show that IOMM not only improves training efficiency but also achieves state-of-the-art (SOTA) performance.

  Implicationmissing

  Implication not extracted yet.

  Verificationpartial

  partial

• Image-Only Training for UMMs (IOMM) is a data-efficient two-stage training framework for Unified Multimodal Models (UMMs) visual generation components.method 70%
  Evidencepartial

  In this paper, we systematically analyze pre-training recipes for UMM visual generation and identify these two issues as the major bottlenecks. To address them, we propose Image-Only Training for UMMs (IOMM), a data-efficient two-stage training framework.

  Implicationpartial

  The abstract explicitly introduces IOMM as a 'data-efficient two-stage training framework' to address bottlenecks in UMM visual generation.

  Verificationpartial

  partial

• The first stage of IOMM pre-trains the visual generative component exclusively using abundant unlabeled image-only data.method 70%
  Evidencepartial

  The first stage pre-trains the visual generative component exclusively using abundant unlabeled image-only data, thereby removing the dependency on paired data for this costly phase.

  Implicationpartial

  The abstract clearly states the purpose and data source for the first stage of the proposed framework.

  Verificationpartial

  partial

• IOMM removes the dependency on paired text-image data for the costly pre-training phase of UMM visual generation.method 70%
  Evidencepartial

  thereby removing the dependency on paired data for this costly phase.

  Implicationpartial

  The abstract directly states this benefit of the first stage of IOMM.

  Verificationpartial

  partial

• The second stage of IOMM fine-tunes the model using a mixture of unlabeled images and a small curated set of text-image pairs.method 70%
  Evidencepartial

  The second stage fine-tunes the model using a mixture of unlabeled images and a small curated set of text-image pairs, leading to improved instruction alignment and generative quality.

  Implicationpartial

  The abstract describes the data used in the second stage of the IOMM framework.

  Verificationpartial

  partial

• The IOMM-B (3.6B) model achieved a GenEval score of 0.89.result 70%
  Evidencepartial

  For example, our IOMM-B (3.6B) model was trained from scratch using only ~1050 H800 GPU hours (with the vast majority, 1000 hours, dedicated to the efficient image-only pre-training stage). It achieves 0.89 on GenEval and 0.55 on WISE--surpassing strong baselines such as BAGEL-7B (0.82 & 0.55) and BLIP3-o-4B (0.84 & 0.50).

  Implicationpartial

  This is a specific, verifiable numerical result presented in the abstract.

  Verificationpartial

  partial