Consistency in Diffusion-Based Visual Generation: A Survey
Abstract
Diffusion models are now a standard foundation for visual generation, including image synthesis, editing, video generation, and 3D-aware content creation. As these models are used for more structured creation, visual quality alone is often insufficient. Generated content can omit prompt details, break compositional relations, lose subject identity across instances, drift over time, disagree across viewpoints, or violate preference, safety, and physical-plausibility requirements. These problems are studied in many task-specific literatures, but they are seldom organized by the agreement relation that a method is designed to enforce. We treat consistency as a requirement on generated content and formulate it as a family of agreement relations. We distinguish three primary relations: external consistency, which concerns agreement with user-specified or task-provided conditions; internal consistency, which concerns coherence among generated parts, views, instances, or temporal states; and normative consistency, which concerns agreement with evaluative criteria such as human preference, safety, and world-level plausibility. We review representative methods under this relation-based view, comparing their scope, preserved properties, and enforcement mechanisms. We also summarize evaluation protocols and benchmarks, examine cross-category trade-offs, and discuss open problems in conflict-aware optimization, interpretable metrics, persistent memory, and long-horizon world-consistent…
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.