Strongly lensed supernovae provide a powerful probe of both astrophysics and cosmology, but in wide-field time-domain surveys their multiple images may remain spatially unresolved and appear as a single blended transient source. We present a photometry-only framework for identifying such systems directly from blended light curves and validate it using real Zwicky Transient Facility (ZTF) data.
Building on a model-independent light-curve framework, we use Bayesian inference to recover time delays and relative magnifications without relying on spatially resolved multiple images or fixed supernova templates. Applying the method to 445 well-converged, spectroscopically confirmed ZTF Type Ia supernovae, we find only one false positive when adopting a conservative threshold of Δt ≥ 12 days, corresponding to a false-positive fraction of approximately 0.22%.
We further apply the framework to a larger, unclassified ZTF transient sample, identifying 20 high-priority candidates for further investigation. In addition, we present a separate morphology-based machine-learning approach to identifying strong-lens systems using imaging data from the Rubin Strong Lensing Challenge. These two approaches explore complementary ways of discovering strong gravitational lenses through temporal and spatial information.