Complete formal verification of neural networks is crucial for their deployment in safety-critical domains. A key bottleneck stems from encoding complexity: traditional methods assign one binary variable per unstable ReLU neuron. We propose the Sign-Absolute Reformulation Theory (SART), which fundamentally breaks the conventional one-to-one mapping between unstable neurons and binary variables by establishing formal reducibility criteria. This allows for finer-grained modeling, where each unstable neuron corresponds on average to fewer than one binary variable, thereby reducing verification complexity at its source. Based on SART, we derive a theoretical lower bound on the number of binary variables required for complete verification and, under the assumption that P ≠ NP , prove that variables in the final layer can be compressed by 50%, while the number of variables in intermediate layers cannot be further reduced. To overcome the apparent “last-layer-only” limitation, we recast verification as a sequential process and, crucially, show that the gain lifts to the entire network: LayerABS, a SART-based progressive tightening verifier, iteratively treats intermediate layers as temporary final layers and propagates tight bounds that shrink the global search space and binary-variable counts. Furthermore, we reveal a structural law influencing verification complexity: when the signs of weights of unstable neurons satisfy numerical symmetry, with positive and negative weights equal or differing by at most one, the worst-case verification complexity achieves the theoretical optimum, offering theoretical guidance for the design of verification-friendly architectures. As a general-purpose underlying encoding, the value of SART is independent of specific algorithms. To comprehensively evaluate its effectiveness, we first evaluate the abstraction-free SART encoding, and then integrate it with abstraction techniques to construct the complete verifier LayerABS and its incomplete variant Incomplete-LayerABS. Across benchmarks, our methods surpass state-of-the-art baselines, validating SART’s practical impact.
Xu et al. (Fri,) studied this question.