Dirk Weissenborn

Papers

Total Citations

364

H-Index

About

Dirk Weissenborn is a prominent researcher specializing in deep learning architectures, computer vision, and natural language processing, with a particular focus on developing efficient attention mechanisms for high-dimensional data. His most influential contribution, the Axial Transformer, introduced a groundbreaking approach to self-attention that dramatically reduces the computational burden of processing images and other multidimensional tensors. By decomposing attention across individual axes rather than attending over entire high-dimensional spaces simultaneously, Weissenborn's axial attention mechanism elegantly balances computational efficiency with model expressiveness — a long-standing challenge in autoregressive modeling. This work has garnered over 364 citations, reflecting its significant uptake across both academic research and applied machine learning communities. The Axial Transformer has since influenced a broad range of downstream architectures in image generation, video understanding, and beyond, establishing Weissenborn as a key figure in the movement toward scalable transformer-based models. His research exemplifies how principled architectural innovations can unlock new capabilities in generative modeling, making sophisticated self-attention mechanisms practical for real-world, high-resolution applications.