UniHetero: Could Generation Enhance Understanding for Vision-Language-Model at Large Data Scale?

• URL: http://arxiv.org/abs/2512.23512v2
• Date: Tue, 30 Dec 2025 13:23:48 GMT
• Title: UniHetero: Could Generation Enhance Understanding for Vision-Language-Model at Large Data Scale?
• Abstract summary: A common assumption in unified vision-language models is that adding generation will naturally strengthen understanding.<n>At 200M+ pretraining samples, generation helps understanding only when it operates at the semantic level.<n>Unified generation-understanding demonstrates a superior scaling trend compared to understanding alone.
• Score: 17.49769409149882
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Vision-language large models are moving toward the unification of visual understanding and visual generation tasks. However, whether generation can enhance understanding is still under-explored on large data scale. In this work, we analysis the unified structure with a concise model, UniHetero, under large-scale pretraining (>200M samples). Our key observations are: (1) Generation can improve understanding, but Only if you generate Semantics, Not Pixels. A common assumption in unified vision-language models is that adding generation will naturally strengthen understanding. However, this is not always true at scale. At 200M+ pretraining samples, generation helps understanding only when it operates at the semantic level, i.e. when the model learns to autoregress high-level visual representations inside the LLM. Once pixel-level objectives (e.g., diffusion losses) directly interfere with the LLM, understanding performance often degrades. (2) Generation reveals a superior Data Scaling trend and higher Data Utilization. Unified generation-understanding demonstrates a superior scaling trend compared to understanding alone, revealing a more effective way to learn vision-only knowledge directive from vision modality rather than captioning to text. (3) Autoregression on Input Embedding is effective to capture visual details. Compared to the commonly-used vision encoder, make visual autoregression on input embedding shows less cumulative error and is modality independent, which can be extend to all modalities. The learned semantic representations capture visual information such as objects, locations, shapes, and colors; further enable pixel-level image generation.

Related papers

• Image Generators are Generalist Vision Learners [71.11980587450198]
  We show that image generation training serves a role similar to language understanding and reasoning pretraining.<n>We introduce Vision Banana, a generalist model built by instruction-tuning Nano Banana Pro (NBP) on a mixture of its original training data and vision task data.<n>Our model achieves SOTA results on a variety of vision tasks involving both 2D and 3D understanding, beating or rivaling zero-shot domain-specialists.
  arXiv  Detail & Related papers  (2026-04-22T08:23:48Z)
• Visual Jigsaw Post-Training Improves MLLMs [58.29961336087896]
  We introduce Visual Jigsaw, a generic self-supervised post-training framework designed to strengthen visual understanding in large language models (MLLMs)<n>Visual Jigsaw is formulated as a general ordering task: visual inputs are partitioned, shuffled, and the model must reconstruct the visual information by producing the correct permutation in natural language.<n>Extensive experiments demonstrate substantial improvements in fine-grained perception, temporal reasoning, and 3D spatial understanding.
  arXiv  Detail & Related papers  (2025-09-29T17:59:57Z)
• Impact of Language Guidance: A Reproducibility Study [0.0]
  Recent advances in self-supervised learning allow us to train huge models without explicit annotation.<n>We use an off-the-shelf image captioning model, BLIP-2, to replace the captions and improve performance.<n>We also devise a new metric to evaluate the semantic capabilities of self-supervised models based on interpretability methods.
  arXiv  Detail & Related papers  (2025-04-10T21:59:13Z)
• TULIP: Towards Unified Language-Image Pretraining [60.99500935831526]
  We introduce T, an open-source, drop-in replacement for existing CLIP-like models.<n>Our method leverages generative data augmentation, enhanced image-image and text-text contrastive learning, and image/text reconstruction regularization to learn fine-grained visual features.<n>Our approach, scaling to over 1B parameters, outperforms existing state-of-the-art (SOTA) models across benchmarks.
  arXiv  Detail & Related papers  (2025-03-19T17:58:57Z)
• OTTER: A Vision-Language-Action Model with Text-Aware Visual Feature Extraction [95.6266030753644]
  Vision-Language-Action (VLA) models aim to predict robotic actions based on visual observations and language instructions.<n>Existing approaches require fine-tuning pre-trained vision-language models (VLMs) as visual and language features are independently fed into downstream policies.<n>We propose OTTER, a novel VLA architecture that leverages existing alignments through explicit, text-aware visual feature extraction.
  arXiv  Detail & Related papers  (2025-03-05T18:44:48Z)
• FiVL: A Framework for Improved Vision-Language Alignment through the Lens of Training, Evaluation and Explainability [10.184567639685321]
  We introduce FiVL, a novel method for constructing datasets designed to train LVLMs for enhanced visual grounding.<n>We present benchmarks to assess the model's ability to use image as substantive evidence.<n>We identify attention heads with the strongest vision-language alignment, enabling explainability on visual-driven hallucinations.
  arXiv  Detail & Related papers  (2024-12-19T09:24:10Z)
• MetaMorph: Multimodal Understanding and Generation via Instruction Tuning [57.35160715164359]
  Visual-Predictive Instruction Tuning (VPiT) is a simple and effective extension to visual instruction tuning.<n>VPiT teaches an LLM to predict discrete text tokens and continuous visual tokens from any input sequence of image and text data.<n>We train our MetaMorph model and achieve competitive performance on both visual understanding and generation.
  arXiv  Detail & Related papers  (2024-12-18T18:58:50Z)
• Critical Data Size of Language Models from a Grokking Perspective [35.029074833552656]
  We formalize the phase transition under the grokking configuration into the Data Efficiency Hypothesis. We show that generalization occurs only when language models reach a critical size. Our results deepen the understanding of language model training, offering a novel perspective on the role of data in the learning mechanism of language models.
  arXiv  Detail & Related papers  (2024-01-19T03:24:36Z)
• Visual Grounding Helps Learn Word Meanings in Low-Data Regimes [47.7950860342515]
  Modern neural language models (LMs) are powerful tools for modeling human sentence production and comprehension. But to achieve these results, LMs must be trained in distinctly un-human-like ways. Do models trained more naturalistically -- with grounded supervision -- exhibit more humanlike language learning? We investigate this question in the context of word learning, a key sub-task in language acquisition.
  arXiv  Detail & Related papers  (2023-10-20T03:33:36Z)
• Unified Language-Vision Pretraining in LLM with Dynamic Discrete Visual Tokenization [52.935150075484074]
  We introduce a well-designed visual tokenizer to translate the non-linguistic image into a sequence of discrete tokens like a foreign language. The resulting visual tokens encompass high-level semantics worthy of a word and also support dynamic sequence length varying from the image. This unification empowers LaVIT to serve as an impressive generalist interface to understand and generate multi-modal content simultaneously.
  arXiv  Detail & Related papers  (2023-09-09T03:01:38Z)
• Visual Programming for Text-to-Image Generation and Evaluation [73.12069620086311]
  We propose two novel interpretable/explainable visual programming frameworks for text-to-image (T2I) generation and evaluation. First, we introduce VPGen, an interpretable step-by-step T2I generation framework that decomposes T2I generation into three steps: object/count generation, layout generation, and image generation. Second, we introduce VPEval, an interpretable and explainable evaluation framework for T2I generation based on visual programming.
  arXiv  Detail & Related papers  (2023-05-24T16:42:17Z)
• Localization vs. Semantics: Visual Representations in Unimodal and Multimodal Models [57.08925810659545]
  We conduct a comparative analysis of the visual representations in existing vision-and-language models and vision-only models. Our empirical observations suggest that vision-and-language models are better at label prediction tasks. We hope our study sheds light on the role of language in visual learning, and serves as an empirical guide for various pretrained models.
  arXiv  Detail & Related papers  (2022-12-01T05:00:18Z)
• Visual Speech Recognition for Multiple Languages in the Wild [64.52593130370757]
  We show that designing better VSR models is equally important to using larger training sets. We propose the addition of prediction-based auxiliary tasks to a VSR model. We show that such model works for different languages and outperforms all previous methods trained on publicly available datasets by a large margin.
  arXiv  Detail & Related papers  (2022-02-26T07:21:00Z)
• Does language help generalization in vision models? [0.0]
  We show that a visual model trained on a very large supervised image dataset (ImageNet-21k) can be as efficient for generalization as its multimodal counterpart (CLIP) When compared to other standard visual or language models, the latent representations of BiT-M were found to be just as "linguistic" as those of CLIP.
  arXiv  Detail & Related papers  (2021-04-16T18:54:14Z)
• Visual Distant Supervision for Scene Graph Generation [66.10579690929623]
  Scene graph models usually require supervised learning on large quantities of labeled data with intensive human annotation. We propose visual distant supervision, a novel paradigm of visual relation learning, which can train scene graph models without any human-labeled data. Comprehensive experimental results show that our distantly supervised model outperforms strong weakly supervised and semi-supervised baselines.
  arXiv  Detail & Related papers  (2021-03-29T06:35:24Z)
• Knowledge-Enhanced Personalized Review Generation with Capsule Graph Neural Network [81.81662828017517]
  We propose a knowledge-enhanced PRG model based on capsule graph neural network(Caps-GNN) Our generation process contains two major steps, namely aspect sequence generation and sentence generation. The incorporated knowledge graph is able to enhance user preference at both aspect and word levels.
  arXiv  Detail & Related papers  (2020-10-04T03:54:40Z)

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.