Eureka-Moments in Transformers: Multi-Step Tasks Reveal Softmax Induced Optimization Problems.
David T. Hoffmann, Simon Schrodi, Jelena Bratulic, Nadine Behrmann, Volker Fischer, Thomas Brox
Browse the full ICML paper archive.
David T. Hoffmann, Simon Schrodi, Jelena Bratulic, Nadine Behrmann, Volker Fischer, Thomas Brox
Browse the full ICML paper archive.