Research map / Probabilistic and scientific modeling / Variational and generative sampling
Purrception: Variational Flow Matching for Vector-Quantized Image Generation
Răzvan-Andrei Matișan, Vincent Tao Hu, Grigory Bartosh, Björn Ommer, Cees G. M. Snoek, Max Welling, Jan-Willem van de Meent, Mohammad Mahdi Derakhshani, Floor Eijkelboom
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 the vq-ds8-c2i tokenizer instead of vq-f8 (ablation) | Purrception DiT-XL/2 · ImageNet-1k 256x256, FID-50k tau=0.8 | FID: 10.85 -> 6.46 | Improved |
| use the vq-ds8-c2i tokenizer instead of vq-f8 (ablation) | Purrception DiT-XL/2 · ImageNet-1k 256x256, FID-50k tau=1.0 | FID: 12.33 -> 9.60 | Improved |
| use the vq-ds8-c2i tokenizer instead of vq-f8 (ablation) | Purrception DiT-XL/2 · ImageNet-1k 256x256, FID-50k tau=0.7 | FID: 11.80 -> 7.44 | Improved |
| lower inference softmax temperature from 1.0 to 0.8 (tokenizer vq-ds8-c2i) (ablation) | Purrception DiT-XL/2 · ImageNet-1k 256x256, FID-50k | FID: 9.60 -> 6.46 | Improved |
| lower inference softmax temperature from 1.0 to 0.9 (tokenizer vq-f8) (ablation) | Purrception DiT-XL/2 · ImageNet-1k 256x256, FID-50k | FID: 12.33 -> 10.82 | Improved |
| add z-loss regularization (1e-5 log^2 Z) to the training loss (ablation) | Purrception DiT-XL/2 · training stability | training loss divergence | Improved |
| use Purrception instead of CFM and CFM-endpoint | DiT-L/2 VQ flow · ImageNet-1k FID-10k convergence | training speed-up: 1.65x faster | Improved |
| use Purrception instead of DFM | DiT-L/2 VQ flow · ImageNet-1k FID-10k convergence | training speed-up: 3.0x faster | Improved |
| use Purrception instead of CFM baselines | DiT-XL/2 VQ flow · ImageNet-1k FID-10k convergence | training speed-up: 2.3x faster | Improved |
| use Purrception instead of DFM | DiT-XL/2 VQ flow · ImageNet-1k FID-10k convergence | training speed-up: 3.5x faster | Improved |
Closest papers by meaning
- Variational Flow Matching for Graph Generation (NeurIPS 2024 · similarity 0.87)
- Variational Rectified Flow Matching (ICML 2025 · similarity 0.87)
- Categorical Flow Matching on Statistical Manifolds (NeurIPS 2024 · similarity 0.86)
- Blockwise Flow Matching: Improving Flow Matching Models For Efficient High-Quality Generation (NeurIPS 2025 · similarity 0.86)
- Rethinking Vector Field Learning for Generative Segmentation (NeurIPS 2026 · similarity 0.86)
- Structured Coupling for Flow Matching (NeurIPS 2026 · similarity 0.86)
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.
