Attention Projection Mixing with Exogenous Anchors

Cross-layer reuse of early attention projections can improve optimization and data efficiency, but it creates a structural conflict: the first layer must simultaneously act as a stable, reusable anchor for all deeper layers and as an effective computational block. We show this ''first-layer tension'' is a hidden limiter of internal-anchor designs. We propose ExoFormer, which resolves the conflict by learning exogenous anchor projections outside the sequential layer stack, decoupling the anchor role from computational refinement. We introduce a unified normalized mixing framework that mixes queries, keys, values, and gate logits using learnable coefficients (exploring coefficient granularities: elementwise/headwise/scalar), and we show that normalizing anchor sources is key to stable reuse. ExoFormer variants consistently outperform their internal-anchor counterparts, and the dynamic variant yields 1.5 downstream accuracy points while matching validation loss using 1.5x fewer tokens than Gated Attention. We explain this efficacy via an Offloading Hypothesis: external anchors preserve essential token identity, allowing layers to specialize exclusively in refinement. We release code and models to facilitate future research.

Attention Projection Mixing and Exogenous Anchors

Transformers that reuse early-layer attention projections as residuals face a fundamental tension: the first layer must simultaneously serve as a stable reference for all deeper layers and as an effective computational block. To resolve this, we propose ExoFormer, which learns dedicated exogenous anchor projections outside the sequential layer stack, decoupling the anchor role from computational refinement. Through a unified normalized mixing framework (studying different coefficient granularities: elementwise, headwise, scalar) across all attention pathways (queries, keys, values, and gate logits), ExoFormer variants consistently outperform their internal-anchor counterparts. Moreover, the dynamic variant achieves a 2.13-point increase in downstream accuracy over the baseline and demonstrates superior data efficiency, matching baseline validation loss with 1.84x fewer tokens. ExoFormer also achieves a 2x reduction in attention sink compared to standard Gated Attention. Paradoxically, all ExoFormer variants exhibit signs of representation collapse. We explain this via an Offloading Hypothesis: external anchors preserve essential token identity, allowing layers to specialize exclusively in computational refinement. We release codes and models to facilitate future research.

AI Enabling Technologies: A Survey

Artificial Intelligence (AI) has the opportunity to revolutionize the way the United States Department of Defense (DoD) and Intelligence Community (IC) address the challenges of evolving threats, data deluge, and rapid courses of action. Developing an end-to-end artificial intelligence system involves parallel development of different pieces that must work together in order to provide capabilities that can be used by decision makers, warfighters and analysts. These pieces include data collection, data conditioning, algorithms, computing, robust artificial intelligence, and human-machine teaming. While much of the popular press today surrounds advances in algorithms and computing, most modern AI systems leverage advances across numerous different fields. Further, while certain components may not be as visible to end-users as others, our experience has shown that each of these interrelated components play a major role in the success or failure of an AI system. This article is meant to highlight many of these technologies that are involved in an end-to-end AI system. The goal of this article is to provide readers with an overview of terminology, technical details and recent highlights from academia, industry and government. Where possible, we indicate relevant resources that can be used for further reading and understanding.

This looks like that: deep learning for interpretable image recognition

When we are faced with challenging image classification tasks, we often explain our reasoning by dissecting the image, and pointing out prototypical aspects of one class or another. The mounting evidence for each of the classes helps us make our final decision. In this work, we introduce a deep network architecture that reasons in a similar way: the network dissects the image by finding prototypical parts, and combines evidence from the prototypes to make a final classification. The algorithm thus reasons in a way that is qualitatively similar to the way ornithologists, physicians, geologists, architects, and others would explain to people on how to solve challenging image classification tasks. The network uses only image-level labels for training, meaning that there are no labels for parts of images. We demonstrate the method on the CIFAR-10 dataset and 10 classes from the CUB-200-2011 dataset.