Research map / Graph learning / Graph self-supervised learning
CheckMate! Watermarking Graph Diffusion Models in Polynomial Time
Roberto Gheda, Abele Mălan, Robert Birke, Maksim Kitsak, Lydia Y. Chen
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 approximate dequantization before detection (ablation) | CheckWate · Planar | Z-score: 0.7 -> 67.6 | Improved |
| add approximate dequantization before detection (ablation) | CheckWate · Proteins | Z-score: 2.0 -> 404.7 | Improved |
| add latent sparsification robust detection (ablation) | CheckWate · Planar (5% edge addition) | Z-score: 0.0 -> 32.6 | Improved |
| embed checkerboard watermark vs Gaussian Shading under isomorphism | CheckWate · Planar | Z-score: 0 -> 41.8 | Improved |
| use CheckWate instead of Gaussian Shading on Proteins | graph diffusion (GruM) · Proteins | Z-score: 110.3 -> 404.7 | Improved |
| use CheckWate watermark instead of none | graph diffusion · Proteins | Degree MMD: 0.4315 -> 0.0473 | Improved |
| use CheckWate watermark instead of none | graph diffusion · Planar | V.U.N. (%): 72.5 -> 67.5 | Hurt |
| use TreeRing watermark | graph diffusion · Planar | V.U.N. (%): 72.5 -> 10 | Hurt |
| use Bipartite watermark | graph diffusion · SBM | V.U.N. (%): 67.5 -> 0 | Hurt |
| use ideal dequantization instead of approximate (ablation) | CheckWate · SBM (20% node deletion) | Z-score: 9.7 -> 10.3 | No clear effect |
| increase attack strength to 20% node deletion | CheckWate · SBM | Z-score: 32.8 -> 9.7 | Hurt |
Closest papers by meaning
- KGMark: A Diffusion Watermark for Knowledge Graphs (ICML 2025 · similarity 0.87)
- Black-Box Forgery Attacks on Semantic Watermarks for Diffusion Models (CVPR 2025 · similarity 0.85)
- WMAdapter: Adding WaterMark Control to Latent Diffusion Models (ICML 2025 · similarity 0.85)
- T2SMark: Balancing Robustness and Diversity in Noise-as-Watermark for Diffusion Models (NeurIPS 2025 · similarity 0.84)
- TimeWak: Temporal Chained-Hashing Watermark for Time Series Data (NeurIPS 2025 · similarity 0.84)
- SleeperMark: Towards Robust Watermark against Fine-Tuning Text-to-image Diffusion Models (CVPR 2025 · similarity 0.84)
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.
