Pith. sign in

REVIEW 9 cited by

CADS: Unleashing the Diversity of Diffusion Models through Condition-Annealed Sampling

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2310.17347 v4 pith:Q6BVK353 submitted 2023-10-26 cs.CV

Seyedmorteza Sadat , Jakob Buhmann , Derek Bradley , Otmar Hilliges , Romann M. Weber This is my paper · ORCID
classification cs.CV
keywords diffusiondiversitymodelssamplingcadsconditioninggenerationcondition-annealed
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
abstract

While conditional diffusion models are known to have good coverage of the data distribution, they still face limitations in output diversity, particularly when sampled with a high classifier-free guidance scale for optimal image quality or when trained on small datasets. We attribute this problem to the role of the conditioning signal in inference and offer an improved sampling strategy for diffusion models that can increase generation diversity, especially at high guidance scales, with minimal loss of sample quality. Our sampling strategy anneals the conditioning signal by adding scheduled, monotonically decreasing Gaussian noise to the conditioning vector during inference to balance diversity and condition alignment. Our Condition-Annealed Diffusion Sampler (CADS) can be used with any pretrained model and sampling algorithm, and we show that it boosts the diversity of diffusion models in various conditional generation tasks. Further, using an existing pretrained diffusion model, CADS achieves a new state-of-the-art FID of 1.70 and 2.31 for class-conditional ImageNet generation at 256$\times$256 and 512$\times$512 respectively.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 9 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Classifier-Free Guidance: From High-Dimensional Analysis to Generalized Guidance Forms

    cs.LG 2025-02 conditional novelty 7.0 of 10

    CFG's distortion of the target distribution vanishes as data dimension grows, and a power-law generalization improves fidelity and diversity in high-dimensional generative models.

  2. Momentum Guidance: Plug-and-Play Guidance for Flow Models

    cs.LG 2026-02 conditional novelty 6.0 of 10

    Momentum Guidance improves flow-model sample quality by extrapolating the current velocity away from an exponential moving average of past velocities, with no extra model evaluations.

  3. Zero-Shot Cyclic Peptide Design via Composable Geometric Constraints

    cs.LG 2025-07 conditional novelty 6.0 of 10

    CP-Composer trains a geometric diffusion model on linear peptides and imposes cyclization constraints at generation time, achieving 38-84% constraint satisfaction across four cyclization strategies.

  4. DIMCIM: A Quantitative Evaluation Framework for Default-mode Diversity and Generalization in Text-to-Image Generative Models

    cs.CV 2025-06 conditional novelty 6.0 of 10

    A new evaluation framework, DIMCIM, measures default-mode diversity and prompted generalization in text-to-image models, finding a scale trade-off and a 0.85 correlation between default diversity and training data diversity.

  5. Boost-and-Skip: A Simple Guidance-Free Diffusion for Minority Generation

    cs.LG 2025-02 conditional novelty 6.0 of 10

    Starting diffusion sampling from variance-boosted noise and skipping early timesteps generates minority samples at guided-method quality with far less compute.

  6. EMAG: Self-Rectifying Diffusion Sampling with Exponential Moving Average Guidance

    cs.CV 2025-12 conditional novelty 5.0 of 10

    EMAG replaces selected attention maps with their exponential moving average during diffusion sampling, reporting +0.46 HPS over CFG on SD3 and composing with APG/CADS.

  7. DP-DocLDM: Differentially Private Document Image Generation using Latent Diffusion Models

    cs.CR 2025-08 reject novelty 5.0 of 10

    DP-DocLDM fine-tunes a latent diffusion model under differential privacy to generate synthetic document images, but the public pretraining set already contains the private benchmark datasets, invalidating the claimed ...

  8. Synthetic Data Augmentation using Pre-trained Diffusion Models for Long-tailed Food Image Classification

    cs.CV 2025-06 conditional novelty 5.0 of 10

    A two-stage diffusion-based data augmentation pipeline with confusing-class negative prompts improves long-tailed food image classification accuracy on Food101-LT and VFN-LT.

  9. How Much To Guide: Revisiting Adaptive Guidance in Classifier-Free Guidance Text-to-Vision Diffusion Models

    cs.CV 2025-06 conditional novelty 4.0 of 10

    Using classifier-free guidance for only the first 30-50% of denoising steps preserves generation quality while saving 20-30% inference time across image and video diffusion models.

Pith tools