InverseNet: Benchmarking Operator Mismatch Calibration Across Compressive Imaging Modalities

State-of-the-art EfficientSCI loses 20.58 dB -- from 35.39 to 14.81 dB -- when its assumed forward operator deviates from the physical truth by just eight parameters. This operator mismatch is the default condition of every deployed compressive imaging system, yet no existing benchmark quantifies it. We introduce InverseNet, the first cross-modality benchmark for operator mismatch in compressive imaging, spanning coded aperture snapshot spectral imaging (CASSI), coded aperture compressive temporal imaging (CACTI), and single-pixel camera (SPC). InverseNet evaluates 12 reconstruction methods under a four-scenario protocol -- ideal (I), mismatched (II), oracle-corrected (III), and blind grid-search calibration (IV) -- across 27 simulated scenes and 9 real hardware captures, totalling over 360 experiments. We discover an inverse performance-robustness relationship: methods achieving the highest ideal PSNR suffer the largest mismatch degradation -- confirming that a mediocre algorithm with a correct forward model outperforms a state-of-the-art network with a wrong one. We further establish an operator-awareness taxonomy: mask-oblivious architectures show zero calibration benefit, while operator-conditioned methods recover 41-90% of mismatch losses. Scenario IV provides a practical calibration baseline via self-supervised objectives, recovering 85-100% of the oracle bound without ground truth. Real hardware experiments on 5 CASSI scenes and 4 CACTI scenes confirm that simulation-derived patterns hold on physical data.