Research map / Image generation / Fast diffusion sampling
Nonparametric Data Attribution for Diffusion Models
Yutian Zhao, Chao Du, Xiaosen Zheng, Tianyu Pang, Min Lin
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 patch-level analytic-score influence (NDA) instead of CLIP cosine similarity | NDA · CIFAR-2 LDS (validation) | LDS (%): 8.54 -> 24.88 | Improved |
| use NDA instead of D-TRAK gradient-based attribution | NDA · CelebA LDS | LDS gap to D-TRAK: gap 5.94/2.92 (val/gen) | Hurt |
| combine original and low-resolution patch influence (gamma=0.75) (ablation) | NDA · CIFAR-2 validation LDS | LDS (%) | Improved |
| use larger patch sizes at higher timesteps (ablation) | NDA · CIFAR-2 validation LDS | LDS (%) | Improved |
| set top-k patches per training image to 100 (ablation) | NDA · CIFAR-2/CelebA validation LDS | LDS (%) | Improved |
| extend aggregation timesteps beyond tau_max=500 (to >=600) (ablation) | NDA · CIFAR-2 validation LDS | LDS (%): drop of ~0.5 at tau_max>=600 | Hurt |
| use high-noise timesteps (t>=600) for attribution (ablation) | NDA · CIFAR-2 validation LDS | LDS (%) | Hurt |
| use NDA instead of CLIP/random for removal in counterfactual retraining | NDA · CIFAR-2 counterfactual | median l2 distance: 6.68 -> 8.43 | Improved |
Closest papers by meaning
- Diffusion Attribution Score: Evaluating Training Data Influence in Diffusion Models (ICLR 2025 · similarity 0.92)
- Training Data Attribution in Diffusion Models via Mirrored Unlearning and Noise-Consistent Skew (NeurIPS 2026 · similarity 0.91)
- Influence Functions for Scalable Data Attribution in Diffusion Models (ICLR 2025 · similarity 0.91)
- Intriguing Properties of Data Attribution on Diffusion Models (ICLR 2024 · similarity 0.90)
- Tracing the Roots: Leveraging Temporal Dynamics in Diffusion Trajectories for Origin Attribution (NeurIPS 2025 · similarity 0.89)
- The Emergence of Reproducibility and Consistency in Diffusion Models (ICML 2024 · similarity 0.89)
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.
