PRISM: Diversifying Dataset Distillation by Decoupling Architectural Priors
This work addresses a bottleneck in dataset distillation for machine learning practitioners by improving synthetic data diversity and generalization, though it is incremental as it builds on multi-teacher approaches.
The paper tackles the problem of dataset distillation inheriting inductive bias from a single teacher model, which reduces intra-class diversity and generalization, by presenting PRISM, a framework that decouples architectural priors using diverse teacher models, resulting in outperforming existing methods on ImageNet-1K and generating data with richer intra-class diversity.
Dataset distillation (DD) promises compact yet faithful synthetic data, but existing approaches often inherit the inductive bias of a single teacher model. As dataset size increases, this bias drives generation toward overly smooth, homogeneous samples, reducing intra-class diversity and limiting generalization. We present PRISM (PRIors from diverse Source Models), a framework that disentangles architectural priors during synthesis. PRISM decouples the logit-matching and regularization objectives, supervising them with different teacher architectures: a primary model for logits and a stochastic subset for batch-normalization (BN) alignment. On ImageNet-1K, PRISM consistently and reproducibly outperforms single-teacher methods (e.g., SRe2L) and recent multi-teacher variants (e.g., G-VBSM) at low- and mid-IPC regimes. The generated data also show significantly richer intra-class diversity, as reflected by a notable drop in cosine similarity between features. We further analyze teacher selection strategies (pre- vs. intra-distillation) and introduce a scalable cross-class batch formation scheme for fast parallel synthesis. Code will be released after the review period.