Research map / Image generation / Fast diffusion sampling
2ndMatch: Finetuning Pruned Diffusion Models via Second-Order Jacobian Matching
Caleb Zheng, Eli Shlizerman
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 second-order Jacobian matching to NP+KD finetuning (ablation) | Diff-Pruned U-Net · CIFAR-10 | FID: 5.05 -> 4.58 | Improved |
| add first-order Jacobian matching on top of NP+KD (ablation) | Diff-Pruned U-Net · CIFAR-10 | FID: 5.05 -> 5.14 | No clear effect |
| add knowledge distillation to noise prediction (ablation) | Diff-Pruned U-Net · CIFAR-10 | FID: 5.29 -> 5.05 | Improved |
| apply 2ndM finetuning instead of Diff-Pruning finetuning | LDM-4 pruned · ImageNet 256 | FID: 10.23 -> 5.68 | Improved |
| apply 2ndM finetuning instead of Diff-Pruning finetuning | DDPM pruned · LSUN-Bedroom 256 | FID: 17.90 -> 9.68 | Improved |
| apply 2ndM finetuning to BK-SDM pruned models | SDM-v1.4 BK-SDM Tiny · MS-COCO 512 | CLIP score: 0.2653 -> 0.2636 | Hurt |
| apply 2ndM finetuning to BK-SDM pruned models | SDM-v1.4 BK-SDM Base · MS-COCO 512 | FID: 15.76 -> 13.84 | Improved |
| use 2ndM at higher pruning ratio | U-ViT · CIFAR-10 (80% pruning) | FID: 6.68 -> 5.01 | Improved |
| add 2ndM to random pruning (ablation) | Random pruning · CIFAR-10 (44% pruning) | FID: 6.22 -> 5.57 | Improved |
| add 2ndM to Diff-Pruning (ablation) | Diff-Pruning · CIFAR-10 (44% pruning) | FID: 5.29 -> 4.58 | Improved |
| use 2ndM instead of Diff-Pruning finetuning | U-ViT pruned · CelebA 64 | FID: 3.57 -> 3.35 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- BK-SDM Base — "BK-SDM: A Lightweight, Fast, and Cheap Version of Stable Diffusion" (ECCV 2024) · better on 3 of 4 shared results
- DeepCache — DeepCache: Accelerating Diffusion Models for Free (CVPR 2024) · better on 1 of 2 shared results
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
- ALTER: All-in-One Layer Pruning and Temporal Expert Routing for Efficient Diffusion Generation (NeurIPS 2025 · similarity 0.90)
- Diffusion Model for Dense Matching (ICLR 2024 · similarity 0.89)
- Attention-Driven Training-Free Efficiency Enhancement of Diffusion Models (CVPR 2024 · similarity 0.88)
- DiP-GO: A Diffusion Pruner via Few-step Gradient Optimization (NeurIPS 2024 · similarity 0.88)
- Inductive Moment Matching (ICML 2025 · similarity 0.88)
- Diff2Flow: Training Flow Matching Models via Diffusion Model Alignment (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.
