A new paper from researchers at Harvard and the University of Illinois at Urbana-Champaign introduces “Explorative Modeling,” or XM, as a third pretraining axis for generative models, alongside parameters and data. The approach increases the number of candidate generations a model explores during training, yielding a 4.1x increase in FLOP efficiency, a 6.2x boost in sample efficiency, and a 47% improvement in parameter efficiency. Efficiency improvements grew from 7% to 36% as training data increased, and from 13% to 23% as model sizes expanded. The XM approach achieved an FID score of 1.43 on ImageNet at 256×256 resolution without guidance, and the model converged roughly 300 times faster than standard training recipes. The framework extends beyond image generation to video generation, natural language processing, and robotics and control tasks, where models matched or exceeded diffusion model baselines while requiring 16 to 256 times fewer inference steps. The paper also positions XM as enabling true end-to-end training and inference without relying on factorization methods. The paper was authored by Alexi Gladstone from UIUC, along with Heng Ji and Yilun Du, with code available publicly at explorative-modeling.github.io and on GitHub.