Research map / Image generation / Fast diffusion sampling
Watch Your Step: Information Injection in Diffusion Models via Shadow Timestep Embedding
An Huang, Junggab Son, Zuobin Xiong
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 |
|---|---|---|---|
| bind MNIST and Fashion-MNIST to shadow timesteps | DDPM · CIFAR-10 | FID: 24.38 -> 21.82 | Improved |
| bind shadow datasets with STE | DDPM · CIFAR-10 | ACC: 73.41% -> 75.65% | No clear effect |
| bind shadow datasets with STE | DDPM · CIFAR-10 | exposure rate: 8.02% -> 9.17% | Hurt |
| bind shadow datasets with STE | DDPM · MNIST | FID: 1.59 -> 1.85 | No clear effect |
| use STE-shadow as backdoor channel versus BadDiffusion | DDPM · CIFAR-10 | FID: 22.53 -> 22.07 | Improved |
| use STE-explicit timesteps instead of VillanDiffusion | DDPM · CIFAR-10 | FID: 25.66 -> 21.82 | Improved |
| use STE watermark versus ROBIN | DDPM · watermark robustness avg | detection accuracy: 0.98 -> 0.89 | Hurt |
| replace DDPM sampler with DDIM in STE (ablation) | STE · CIFAR-10 | FID: 21.82 -> 22.27 | Hurt |
| replace DDPM sampler with DPM-Solver in STE (ablation) | STE · CIFAR-10 | FID: 21.82 -> 37.88 | Hurt |
| replace DDPM sampler with DPM-Solver in STE (ablation) | STE · MNIST | FID: 1.85 -> 16.14 | Hurt |
| modify scheduler timestep range instead of extending model embedding (ablation) | STE reverse design · multi-dataset generation | sample quality | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 3 on most shared results.
- VillanDiffusion — VillanDiffusion: A Unified Backdoor Attack Framework for Diffusion Models (NeurIPS 2023) · better on 4 of 4 shared results
- SleeperMark — SleeperMark: Towards Robust Watermark against Fine-Tuning Text-to-image Diffusion Models (CVPR 2025) · better on 0 of 1 shared result
- ROBIN — ROBIN: Robust and Invisible Watermarks for Diffusion Models with Adversarial Optimization (NeurIPS 2024) · better on 0 of 1 shared result
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
- Adaptive Non-Uniform Timestep Sampling for Accelerating Diffusion Model Training (CVPR 2025 · similarity 0.88)
- Improving Training Efficiency of Diffusion Models via Multi-Stage Framework and Tailored Multi-Decoder Architecture (CVPR 2024 · similarity 0.86)
- Denoising Task Difficulty-based Curriculum for Training Diffusion Models (ICLR 2025 · similarity 0.86)
- Why DiT Models Underperform as Representation Learners without Long Skip Connections (NeurIPS 2026 · similarity 0.85)
- Timestep Embedding Tells: It's Time to Cache for Video Diffusion Model (CVPR 2025 · similarity 0.85)
- Tracing the Roots: Leveraging Temporal Dynamics in Diffusion Trajectories for Origin Attribution (NeurIPS 2025 · 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.
