SAVIER: Source-Anchored Visual Inspection and Evidence-Guided Repair for Vietnamese Cultural Text-to-Image Generation
Organizations: University of Science, Ho Chi Minh City, Vietnam · Vietnam National University, Ho Chi Minh City, Vietnam
Abstract
We present SAVIER, a training-free, one-pass framework that repairs the depiction of a named Vietnamese cultural entity in a generated image while preserving the scene details requested by the user. The problem is hard because a plausible scene can miss the cues that identify the entity, and prompt-level refinement never inspects the image it produces. SAVIER starts from a Culture-TRIP expanded prompt; a Curator separates source-linked cultural cues from prompt-preservation requirements, a prompt-blind Observer describes the first image, and a Refiner turns the discrepancies into at most three positive instructions before one regeneration with the same seed and a fixed bank of reference photographs. On the 50 bilingual prompts of ViCuP, evaluated with SDXL and FLUX.1-dev against a frozen, human-verified visual contract, SAVIER raises contract fulfillment from 55.6 to 69.6 and from 59.1 to 67.3 over Culture-TRIP and raises CAIRE cultural relevance on FLUX.1-dev. A reference-only ablation that removes the agent repair shows that SAVIER's photographs account for most of the contract gain; the agent roles add country-level relevance on FLUX.1-dev without lowering adherence to the original prompt under the official VQAScore. Project page: https://oxyzgiahuy.github.io/savier/
Figures & tables
| Configuration | CAIRE-VN | CAIRE-CN | VQA | VCFS | CCR |
|---|---|---|---|---|---|
| SDXL | 2.66 | 2.46 | 0.861 | 52.9 | 20.8 |
| SDXL + Culture-TRIP | 3.00 | 2.54 | 0.791 | 55.6 | 15.3 |
| SDXL + SAVIER | 3.36 | 2.36 | 0.805 | 69.6 | 15.2 |
| FLUX | 2.50 | 2.54 | 0.832 | 44.8 | 13.3 |
| FLUX + Culture-TRIP | 2.50 | 3.08 | 0.839 | 59.1 | 15.0 |
| FLUX + SAVIER | 3.00 | 2.88 | 0.833 | 67.3 | 14.2 |
T2I Generation
Text-to-image generation creates images from natural-language prompts using generative models, with research focused on fidelity, control, efficiency, and safety.
See all 332 papers