Research map / Image generation / Autoregressive and latent generation
GROW: Watermark Generation with Progressive Guidance for Diffusion Models
Pengcheng Luo, Zexi Jia, Yijia Zhong, Jinchao Zhang, Jie Zhou
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 |
|---|---|---|---|
| use progressive frequency-domain guidance instead of one-step injection into the final latent (ablation) | GROW · MS-COCO | FID: 157.8 -> 12.32 | Improved |
| use progressive guidance instead of one-step injection (ablation) | GROW · MS-COCO | PSNR: 11.36 -> 27.54 | Improved |
| increase watermark strength alpha from 0.3 to 0.7 (ablation) | GROW · average M-ACC | M-ACC: M-ACC 92% -> 99%; FID 9.5 -> 28.2 | Mixed |
| increase guidance scale eta from 50 to 100 (ablation) | GROW · average M-ACC | M-ACC: 90% -> 99% | Improved |
| increase guidance scale eta beyond 100 to 125 (ablation) | GROW · average M-ACC | FID: FID 15.1 at eta=125 | No clear effect |
| start guidance earlier (r_start 0.3) (ablation) | GROW · average M-ACC | M-ACC: 97% at r_start = 0.3 | Improved |
| start guidance late (r_start 0.9) (ablation) | GROW · average M-ACC | M-ACC: 92% at r_start = 0.9 | Hurt |
| replace initial-noise watermark with progressive guidance in extraction (no DDIM inversion) | GROW vs Tree-Ring/WIND · Extraction time on T4 | extraction time (s): 19.70 -> 0.24 | Improved |
| use GROW instead of initial-noise watermarking on SDXL and SD v1.5 | GROW · SDXL base 1.0 | average M-ACC: 0.928 -> 0.981 vs WIND | Improved |
Closest papers by meaning
- Shallow Diffuse: Robust and Invisible Watermarking through Low-Dim Subspaces in Diffusion Models (NeurIPS 2025 · similarity 0.90)
- CoSDA: Enhancing the Robustness of Inversion-based Generative Image Watermarking Framework (AAAI 2025 · similarity 0.89)
- T2SMark: Balancing Robustness and Diversity in Noise-as-Watermark for Diffusion Models (NeurIPS 2025 · similarity 0.89)
- Secure Seed-Based Multi-bit Watermarking for Diffusion Models from First Principles (NeurIPS 2026 · similarity 0.89)
- ROBIN: Robust and Invisible Watermarks for Diffusion Models with Adversarial Optimization (NeurIPS 2024 · similarity 0.88)
- Harnessing Frequency Spectrum Insights for Image Copyright Protection Against Diffusion Models (CVPR 2025 · 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.
