Paper 2602.02437
UniReason 1.0: A Unified Reasoning Framework for World Knowledge Aligned Image Generation and Editing
- Published
- Feb 2026
- Research lab
- Independent
- Citations
- 6
- GitHub
- 144 stars
01 In brief
Summary
UniReason is a unified framework for text-to-image (T2I) generation and image editing, built on the Bagel architecture.
It addresses limitations of existing methods by incorporating world knowledge-enhanced textual reasoning before synthesis and fine-grained editing-like visual refinement after initial generation.
The framework uses two complementary reasoning paradigms: textual reasoning infers implicit knowledge across five categories (cultural commonsense, natural science, spatial, temporal, logical), while visual refinement corrects errors via self-reflection, structurally analogous to image editing.
Training data includes ~300k samples for reasoning and an agent-generated corpus for refinement.
A two-stage training strategy first strengthens generation, then jointly trains understanding and generation branches.
Experiments show state-of-the-art performance among open-source models on WISE, KrisBench, and UniREditBench, with competitive results against closed-source models like GPT-4o and Seedream 4.0.
Ablations confirm the contribution of each component, and a correlation analysis shows that higher editing capability leads to greater refinement gains.
02 From the paper
Abstract
Unified multimodal models often struggle with complex synthesis tasks that demand deep reasoning, and typically treat text-to-image generation and image editing as isolated capabilities rather than interconnected reasoning steps. To address this, we propose UniReason, a unified framework that harmonizes these two tasks through two complementary reasoning paradigms. We incorporate world knowledge-enhanced textual reasoning into generation to infer implicit knowledge, and leverage editing capabilities for fine-grained editing-like visual refinement to further correct visual errors via self-reflection. This approach unifies generation and editing within a shared architecture, mirroring the human cognitive process of planning followed by refinement. We support this framework by systematically constructing a large-scale reasoning-centric dataset (~300k samples) covering five major knowledge domains (e.g., cultural commonsense, physics, etc.) for textual reasoning, alongside an agent-generated corpus for visual refinement. Extensive experiments demonstrate that UniReason achieves advanced performance on reasoning-intensive benchmarks such as WISE, KrisBench and UniREditBench, while maintaining superior general synthesis capabilities.