Bringing together researchers across levels at the intersection of theory, experimental data, and machine learning to explore a paradigm shift in understanding sensory processing: Adaptive and Flexible Representations.  We will discuss how the efficient coding framework operates in dynamic, context-dependent settings, while critically examining the limitations of classical theories for understanding vision across brains and machines.

Vision is remarkably efficient, but the natural world it operates within is rarely stationary. To understand how biological and artificial systems process rich visual detail under constant uncertainty, we must move beyond classical assumptions of static optimization. This workshop brings together leading researchers to investigate adaptive and flexible representations—how sensory systems dynamically adjust their coding strategies to changing contexts, environments, and tasks.

Two primary goals drive the workshop:

• Expanding the Framework: We will explore new ways the efficient coding hypothesis is being used to study dynamic and context-dependent brain and behavioral mechanisms. This includes modern approaches to predictive coding, rapid gain control, and Bayesian uncertainty.

• Addressing the Limitations: We will critically discuss the boundaries and shortcomings of the classical Barlow framework, tackling paradoxes like cortical overcompleteness and complex contextual modulation.

By integrating theory, machine learning, and experimental data, this interdisciplinary workshop aims to advance our conceptualization of vision as a highly flexible, continuously adapting process. Achieving a deeper understanding of these mechanisms has profound implications for both neuroscience and the development of robust, next-generation computer vision systems.