Research map / Image generation / Text-to-image diffusion
Where Culture Fades: Revealing the Cultural Gap in Text-to-Image Generation
Chuancheng Shi, Shangze Li, Shiming Guo, Simiao Xie, Wenhua Wu, Jingtong Dou, Chao Wu, Canran Xiao, Cong Wang, Zifeng Cheng, Fei Shen, Tat-Seng Chua
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add culture-modifier to noun-only prompt | AltDiffusion · CultureBench test (culture-modifier + noun prompts) | CultureVQA: 44.39 | Improved |
| add culture-modifier to noun-only prompt | PEA-Diffusion · CultureBench test (culture-modifier + noun prompts) | CultureVQA: 35.62 | Improved |
| add zero-training neuron amplification (targeted neurons) | PEA-Diffusion · CultureBench test (noun-only prompts) | CultureVQA: 21.65 -> 33.91 | Improved |
| add fine-tuned layer enhancer | PEA-Diffusion · CultureBench test (noun-only prompts) | CultureVQA: 21.65 -> 36.63 | Improved |
| add zero-training neuron amplification (targeted neurons) | AltDiffusion · CultureBench test (noun-only prompts) | CultureVQA: 23.05 -> 30.06 | Improved |
| add fine-tuned layer enhancer | AltDiffusion · CultureBench test (noun-only prompts) | CultureVQA: 23.05 -> 32.66 | Improved |
| amplify random neurons instead of identified ones (ablation) | AltDiffusion · CultureBench test (noun-only prompts) | CultureVQA: 23.05 -> 20.38 | Hurt |
| amplify random neurons instead of identified ones (ablation) | PEA-Diffusion · CultureBench test (noun-only prompts) | CultureVQA: 21.65 -> 21.04 | No clear effect |
| fine-tune random layer instead of culture layer (ablation) | PEA-Diffusion · CultureBench test (noun-only prompts) | CultureVQA: 21.65 -> 22.34 | No clear effect |
| mask identified top-K culture neurons (vs not masking) (ablation) | PEA-Diffusion · CultureBench test (culture-modifier + noun prompts) | CultureVQA: 35.62 -> 7.65 | Improved |
| mask random neurons (vs not masking) (ablation) | PEA-Diffusion · CultureBench test (culture-modifier + noun prompts) | CultureVQA: 35.62 -> 33.04 | No clear effect |
| mask identified top-K culture neurons (vs not masking) (ablation) | AltDiffusion · CultureBench test (culture-modifier + noun prompts) | CultureVQA: 44.54 -> 12.04 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- AltDiffusion — AltDiffusion: A Multilingual Text-to-Image Diffusion Model (AAAI 2024) · better on 4 of 4 shared results
- PEA-Diffusion — PEA-Diffusion: Parameter-Efficient Adapter with Knowledge Distillation in non-English Text-to-Image Generation (ECCV 2024) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- See It from My Perspective: How Language Affects Cultural Bias in Image Understanding (ICLR 2025 · similarity 0.90)
- Beyond Aesthetics: Cultural Competence in Text-to-Image Models (NeurIPS 2024 · similarity 0.89)
- AltDiffusion: A Multilingual Text-to-Image Diffusion Model (AAAI 2024 · similarity 0.88)
- MuLan: Adapting Multilingual Diffusion Models for Hundreds of Languages with Negligible Cost (ICML 2025 · similarity 0.88)
- No Filter: Cultural and Socioeconomic Diversity in Contrastive Vision-Language Models (NeurIPS 2024 · similarity 0.88)
- Culture In a Frame: C$^3$B as a Comic-Based Benchmark for Multimodal Culturally Awareness (ICLR 2026 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
