Research map / Image generation / Fast diffusion sampling
MUSE: Model-Agnostic Tabular Watermarking via Multi-Sample Selection
Liancheng Fang, Aiwei Liu, Henry Peng Zou, Yankai Chen, Hengrui Zhang, Zhongfen Deng, Philip S. Yu
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 Bernoulli(0.5) score instead of uniform score (ablation) | MUSE · Adult (100 rows) | z-statistic: 5.012 -> 7.348 | Improved |
| use adaptive instead of fixed column selection (ablation) | MUSE · Adult (100 rows) | z-statistic: 5.439 -> 7.348 | Improved |
| use adaptive instead of fixed column selection (ablation) | MUSE · Adult (100 rows) | C2ST fidelity: 0.601 -> 0.883 | Improved |
| add masking of previously watermarked values (ablation) | MUSE · Adult (100 rows) | C2ST fidelity: 0.883 -> 0.940 | Improved |
| add masking of previously watermarked values (ablation) | MUSE · Adult (100 rows) | z-statistic: 7.348 -> 4.819 | Hurt |
| increase selected columns from 1 to 3 (ablation) | MUSE · Adult (100 rows) | z-statistic: 4.987 -> 7.348 | Improved |
| increase selected columns from 3 to 5 (ablation) | MUSE · Adult (100 rows) | z-statistic: 7.348 -> 8.624 | Improved |
| increase selected columns from 5 to 7 (ablation) | MUSE · Adult (100 rows) | z-statistic: 8.624 -> 8.728 | No clear effect |
| increase selected columns from 1 to 3 (ablation) | MUSE · Adult (100 rows) | C2ST fidelity: 0.544 -> 0.883 | Improved |
| apply MUSE to autoregressive DP-TBART instead of TabSyn (ablation) | MUSE · Adult (100 rows) | z-statistic: 7.348 -> 7.544 | No clear effect |
| apply MUSE to masked model TabDAR instead of TabSyn (ablation) | MUSE · Adult (100 rows) | z-statistic: 7.348 -> 7.270 | No clear effect |
| add MUSE over TabWak watermarking | TabWak · Adult (100 rows) | detection AUC: 0.844 -> 1.000 | Improved |
Closest papers by meaning
- BitMark: Watermarking Bitwise Autoregressive Image Generative Models (NeurIPS 2025 · similarity 0.86)
- SleeperMark: Towards Robust Watermark against Fine-Tuning Text-to-image Diffusion Models (CVPR 2025 · similarity 0.86)
- Break the Trade-off Between Watermark Strength and Speculative Sampling Efficiency for Language Models (ICLR 2026 · similarity 0.86)
- T2SMark: Balancing Robustness and Diversity in Noise-as-Watermark for Diffusion Models (NeurIPS 2025 · similarity 0.86)
- Shallow Diffuse: Robust and Invisible Watermarking through Low-Dim Subspaces in Diffusion Models (NeurIPS 2025 · similarity 0.85)
- Inevitable Trade-off between Watermark Strength and Speculative Sampling Efficiency for Language Models (NeurIPS 2024 · similarity 0.85)
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.
