We present a transient GENERIC Informed Neural Network (GINN) framework for modeling and validation of complex fluids using transient rheological protocols. The steady-state GINN methodology employs steady-state experimental protocols to train a neural network satisfying the GENERIC thermodynamic equation form [1]. This encompasses standard rheometry such as steady shear flow, where velocity and strain-rate are prescribed, as well as complex flows where fluid particles experience mixed shear and extensional conditions under a mixed velocity field profile. These protocols enable acquisition of microstructure conformation tensor and stress tensor data for GINN training. Extending to transient protocols enables exploration of a broader rheological parameter space, thereby improving and expanding the applicability of GINN models. To investigate the workflow with transient data, we selected canonical rheological protocols: Small Amplitude Oscillatory Shear (SAOS) and Large Amplitude Oscillatory Shear (LAOS). The enhanced GINN training is then used to validate two complementary workflows for an Oldroyd-B type fluid: GINN trained with steady and transient rheometry protocols (simple shear, SAOS, and LAOS) versus training performed exclusively with complex-flow data. Both training approaches are evaluated for their ability to reproduce complex flows and standard rheometry cases. The learned constitutive model is geometry-agnostic and enables application to flows with different topologies than those used for training. This work establishes transient GINN training as a powerful strategy for constitutive discovery in complex rheological flows. REFERENCES [1] Nieto Simavilla D, Bonfanti A, García-Beristain I, Español P, Ellero M. Hammering at the entropy: a GENERIC-guided approach to learning polymeric rheological constitutive equations using PINNs. Journal of Fluid Mechanics. 2025;1016:A11. doi:10.1017/jfm.2025.10325