Paper 2604.28185
Visual Generation in the New Era: An Evolution from Atomic Mapping to Agentic World Modeling
- Published
- Apr 2026
- Research lab
- Independent
- Citations
- 4
- GitHub
- 128 stars
01 In brief
Summary
This roadmap paper argues that visual generation must evolve from appearance synthesis to intelligent visual generation, grounded in structure, dynamics, and causal relations.
It proposes a five-level taxonomy—Atomic, Conditional, In-Context, Agentic, and World-Modeling Generation—to organize progress from passive rendering to interactive, world-aware systems.
The paper analyzes key technical drivers, including diffusion-to-flow matching, unified understanding-and-generation models, and post-training alignment, and reviews data, infrastructure, and applications.
It critiques current benchmarks for overestimating progress and introduces in-the-wild stress tests that reveal failures in spatial logic, physical reasoning, and identity preservation.
The roadmap outlines future directions such as visual chain-of-thought, closed-loop agents, tool-augmented rendering, and world simulation, emphasizing the need for evaluation protocols that test structural and causal correctness.
02 From the paper
Abstract
Recent visual generation models have made major progress in photorealism, typography, instruction following, and interactive editing, yet they still struggle with spatial reasoning, persistent state, long-horizon consistency, and causal understanding. We argue that the field should move beyond appearance synthesis toward intelligent visual generation: plausible visuals grounded in structure, dynamics, domain knowledge, and causal relations. To frame this shift, we introduce a five-level taxonomy: Atomic Generation, Conditional Generation, In-Context Generation, Agentic Generation, and World-Modeling Generation, progressing from passive renderers to interactive, agentic, world-aware generators. We analyze key technical drivers, including flow matching, unified understanding-and-generation models, improved visual representations, post-training, reward modeling, data curation, synthetic data distillation, and sampling acceleration. We further show that current evaluations often overestimate progress by emphasizing perceptual quality while missing structural, temporal, and causal failures. By combining benchmark review, in-the-wild stress tests, and expert-constrained case studies, this roadmap offers a capability-centered lens for understanding, evaluating, and advancing the next generation of intelligent visual generation systems.