From Corpora to Co-Evolving Capabilities: Capability-Centric Data Design for Generalist Image Generation
Summary
The paper introduces a capability-driven data infrastructure with curriculum scheduling to train generalist image generation models using heterogeneous supervision for diverse generative tasks.
View Cached Full Text
Cached at: 08/19/26, 03:57 AM
Paper page - From Corpora to Co-Evolving Capabilities: Capability-Centric Data Design for Generalist Image Generation
Source: https://huggingface.co/papers/2608.18076 Authors:
,
,
,
,
,
,
,
,
,
,
,
,
,
,
,
Abstract
A capability-driven data infrastructure with curriculum scheduling and specialized data engines trains large multimodal diffusion models on curated heterogeneous supervision for diverse generative tasks.
Large-scale image generation has benefited from advances in data scale, quality, rebalancing, and recaptioning, yet conventional pipelines typically optimize task-specific datasets in isolation. A central challenge is not only how to curate each task-specific corpus, but also how to organize heterogeneous supervision according to the dependencies among generative capabilities. We present acapability-driven data infrastructurethat couples capability-specific supervision construction with capability-alignedcurriculum scheduling. Its three specialized yet interoperable data engines build complementaryrelational supervisionfortext-image grounding,inter-image transformation, andimage-knowledge association, whilecaption expertsalignT2Iand editing supervision across tasks and granularities. Amulti-stage curriculumjointly evolves task composition, visual-concept distribution, data quality, and image resolution along the dependency order of capability acquisition, withcapability-aware evaluationclosing the loop through targeted retrieval, expert construction, and gap-aware resampling. At scale, the framework curates a 440M-imageT2Icorpus, 120M editing pairs, and over 27M image-entity pairs. With this infrastructure, we trainmultimodal diffusion modelsat two scales from scratch, with 3B and 6B sizes respectively. We conduct quantitative evaluation onCPI-Bench, along with qualitative evaluations across diverse text-to-image and editing scenarios. Experimental results present broad visual coverage, versatile rendering, and effective transfer across generative capabilities.
View arXiv pageView PDFAdd to collection
Get this paper in your agent:
hf papers read 2608\.18076
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2608.18076 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2608.18076 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2608.18076 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
Search Beyond What Can Be Taught: Evolving the Knowledge Boundary in Agentic Visual Generation
This paper addresses the knowledge boundary problem in visual generation by introducing the SearchGen-20K benchmark and SearchGen-Corpus-1M, and proposes a teach-then-search co-training framework to handle evolving, long-tailed user requests beyond a generator's training data.
Video Generators as General-Purpose Vision Models (8 minute read)
GenCeption repurposes pre-trained video generative models into a single unified feed-forward vision model that achieves state-of-the-art performance across multiple tasks with exceptional data efficiency, marking a shift toward general-purpose visual intelligence.
Post-Generation Curation of Synthetic Images via Homogeneous-Heterogeneous Splitting
This paper proposes a generator-agnostic post-generation curation method that selects informative subsets of synthetic images by splitting real classes into canonical homogeneous and non-redundant heterogeneous subsets, and scoring synthetic images via a fidelity-diversity criterion. It consistently outperforms existing data-selection baselines and matches real-data performance with up to 40% fewer synthetic samples.
Video Generation Models are General-Purpose Vision Learners
This paper proposes that large-scale text-to-video generation can serve as a powerful pre-training paradigm for computer vision, introducing GenCeption which achieves state-of-the-art performance across diverse vision tasks with high data efficiency and emergent generalization to unseen domains.
SCOPE: Structured Decomposition and Conditional Skill Orchestration for Complex Image Generation
SCOPE is a specification-guided framework for text-to-image generation that tracks semantic commitments to better fulfill complex visual intents. It introduces the Gen-Arena benchmark and demonstrates strong performance on complex generation tasks.