Tag
The paper identifies two causes why logic gate networks fail to benefit from increased depth and proposes Input-Anchored Logic Gate Networks (IALGNs) that condition each layer on original inputs, achieving consistent depth-accuracy improvements beyond 100 layers.
Researchers from MIT present a methodology for inverse design of nuclear critical experiments using deep neural networks with a novel multigroup attention pooling architecture and gradient-based optimization to maximize neutronic similarity coefficients. The approach is applied to validate a HALEU fuel transportation cask, achieving high similarity scores for three configurations of interest.
LeapAlign is a post-training method that improves flow matching model alignment with human preferences by reducing computational costs through two-step trajectory shortcuts while enabling stable gradient propagation to early generation steps. The method outperforms state-of-the-art approaches when fine-tuning Flux models across various image quality and text-alignment metrics.