Ali Kayyam

2 accepted papers

2026

Do Transformers Need Three Projections? Systematic Study of QKV Variants

ICML 2026poster

Transformers have become the standard solution for various AI tasks, with the query, key, and value (QKV) attention formulation playing a central role. However, the individual contribution of these three projections and the impact of omitting some remain poorly understood. We systematically evaluate…

Cited by 0 Source ↗Scholar ↗

2026

Position: Evidence and Implications of Texture Bias in Deep Neural Networks

Ali Kayyam

ICML 2026poster

Whether deep vision models recognize objects primarily by shape or texture remains a central and unresolved question in computer vision. Early studies report a strong texture bias in convolutional neural networks (CNNs), while other work reports shape-biased representations. We argue that much of th…

Cited by 0 Source ↗Scholar ↗