PULVINAR PATHWAYS AS SKIP CONNECTIONS IN DEEP NEURAL NETWORKS FOR VISION

Abstract
The pulvinar is thought to regulate cortico-cortical communication through transthalamic pathways, yet the computational consequences of such modulation remain unclear. Here we test a minimal pulvinar-inspired long-range skip pathway in a hierarchical convolutional vision model, designed as a learned projection from early to late feature maps coupled to a gain-controlled gating mechanism. We evaluated the architecture in two complementary experiments. First, in CIFAR-10 categorization, the pulvinar-augmented model did not primarily act as a uniform accuracy booster; instead, it reshaped contrast-dependent response scaling and stabilized internal representations in a way consistent with gain control. Second, in a near-threshold contrast-detection paradigm with noisy backgrounds, pulvinar-like modulation systematically redistributed detection outcomes, consistent with a change in decision criterion under uncertainty, and in specific regimes supported improved discriminability. Together, these results indicate that pulvinar-inspired connectivity contributes most reliably to gain/criterion control. Thus, we suggest that the pulvinar regulates how evidence is converted into decisions, rather than to raw task accuracy.