Abstract
Generative models often suffer from mode collapse and limited sample diversity. While prior works attempt to mitigate this by jointly generating a batch of samples and repelling their trajectories, these heuristics do not explicitly maximize the diversity of the resulting endpoints. We introduce JIVE, a training-free framework that enhances generative diversity by injecting velocity perturbations aligned with the leading right singular subspace of the generator's endpoint Jacobian. By leveraging this local geometric structure, JIVE provably maximizes endpoint diversity while preserving sample quality. To maintain practical efficiency, we compute these perturbation directions via matrix-free iterations rooted in classical numerical linear algebra, requiring only a small computational overhead. Across different benchmarks, JIVE boosts both pixel and feature-level diversity in few-step and one-step generation.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Zhang, G., Cai, B., Zhang, B., Chen, C., & Tang, R. (2026). JIVE: Jacobian-Informed Volume Expansion for Diverse Generative Sampling. https://omanscience.com/en/articles/jive-jacobian-informed-volume-expansion-for-diverse-generative-sampling
MLA 9
Zhang, Guangxun, et al. "JIVE: Jacobian-Informed Volume Expansion for Diverse Generative Sampling." https://omanscience.com/en/articles/jive-jacobian-informed-volume-expansion-for-diverse-generative-sampling.
Chicago (author–date)
Zhang, Guangxun, Brian Cai, Boxuan Zhang, Chao Chen, and Ruixiang Tang. 2026. "JIVE: Jacobian-Informed Volume Expansion for Diverse Generative Sampling." https://omanscience.com/en/articles/jive-jacobian-informed-volume-expansion-for-diverse-generative-sampling.
Harvard
Zhang, G., Cai, B., Zhang, B., Chen, C. and Tang, R. (2026) 'JIVE: Jacobian-Informed Volume Expansion for Diverse Generative Sampling', Available at: https://omanscience.com/en/articles/jive-jacobian-informed-volume-expansion-for-diverse-generative-sampling.
Vancouver
Zhang G, Cai B, Zhang B, Chen C, Tang R. JIVE: Jacobian-Informed Volume Expansion for Diverse Generative Sampling. https://omanscience.com/en/articles/jive-jacobian-informed-volume-expansion-for-diverse-generative-sampling
IEEE
G. Zhang, B. Cai, B. Zhang, C. Chen, and R. Tang, "JIVE: Jacobian-Informed Volume Expansion for Diverse Generative Sampling," https://omanscience.com/en/articles/jive-jacobian-informed-volume-expansion-for-diverse-generative-sampling.