Computational evolutionary perception

Hoffman and Singh raised the multi-task question explicitly and conjectured that general-purpose encodings would not become more veridical with more tasks, calling for detailed mathematical models. dalla2026between supplies those models and gives the opposite answer under a separation condition.