Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2603.00408

Exact and Asymptotically Complete Robust Verifications of Neural Networks via Ising Solvers

Wenxin Li, Wenchao Liu, Weihao Li, Chuan Wang, Qi Gao, Yin Ma, Hai Wei, Kai Wen

cs.LGcs.AIphysics.opticsquant-phSubmitted Feb 28, 2026• Announced Mar 3, 2026• Revised Aug 11, 2026 (v4)
v1 · Feb 28, 2026v2 · Jul 27, 2026v3 · Jul 30, 2026v4 · Aug 11, 2026v5 · Oct 22, 2026
arXivHTMLPDF

Abstract

We present an Ising-compatible framework for formal neural-network robustness verification under bounded input perturbations. For piecewise-linear activations, the Exact Logarithmic PWL Model (Log-PWL) provides an exact, sound, and complete formulation with a state-optimal logarithmic encoding, reducing the binary variables per neuron from linear to information-theoretically minimal logarithmic complexity. For general bounded element-wise activations, the Asymptotic Step-Envelope Model (Step-Env) uses sound piecewise-constant envelopes whose lower and upper neuron states remain decision variables coupled to a common adversarial input. We prove that its globally optimized output bounds converge uniformly to the true network extrema as the segment width vanishes, yielding asymptotic completeness of verification. We further develop a hybrid Benders solver with output-sensitive iteration bounds, leveraging a Helly-based conflict compression theory to avoid generic exponential cut explosion. Interval pruning, certificate transfer for pruned networks, and layerwise classical--Ising partitioning further reduce spin requirements. Experiments show exact certification fidelity for piecewise-linear networks and near-reference accuracy for sigmoid networks with compact spin budgets.

Related papers

Quantum Robustness Verification: A Hybrid Quantum-Classical Neural Network Certification Algorithm

Nicola Franco, Tom Wollschläger, Nicholas Gao +2 more

In recent years, quantum computers and algorithms have made significant progress indicating the prospective importance of quantum computing (QC). Especially combinatorial optimization has gained a lot of attention as an application field for near-term quantum computers, both by using gate-based QC via the Quantum Approximate Optimization Algorithm and by quantum annealing using the Ising model. However, demonstrating an advantage over classical methods in real-world applications remains an active area of research. In this work, we investigate the robustness verification of ReLU networks, which involves solving many-variable mixed-integer programs (MIPs), as a practical application. Classically, complete verification techniques struggle with large networks as the combinatorial space grows exponentially, implying that realistic networks are difficult to be verified by classical methods. To alleviate this issue, we propose to use QC for neural network verification and introduce a hybrid quantum procedure to compute provable certificates. By applying Benders decomposition, we split the MIP into a quadratic unconstrained binary optimization and a linear program which are solved by quantum and classical computers, respectively. We further improve existing hybrid methods based on the Benders decomposition by reducing the overall number of iterations and placing a limit on the maximum number of qubits required. We show that, in a simulated environment, our certificate is sound, and provides bounds on the minimum number of qubits necessary to approximate the problem. Finally, we evaluate our method within simulations and on quantum hardware.

arXiv:2205.00900•IEEE International Conference on Quantum Computing and Engineering (QCE)•Sep 1, 2022•29 citations

Robustness Verification of Binary Neural Networks: An Ising and Quantum-Inspired Framework

Rahul Kumar Singh, Seyran Saeedi, Zheng Zhang

Binary neural networks (BNNs) are increasingly deployed in edge computing applications due to their low hardware complexity and high energy efficiency. However, verifying the robustness of BNNs against input perturbations, including adversarial attacks, remains computationally challenging because the underlying decision problem is inherently combinatorial. In this paper, we propose an Ising- and quantum-inspired framework for BNN robustness verification. We show that, for a broad class of BNN architectures, robustness verification can be formulated as a Quadratic Constrained Boolean Optimization (QCBO) problem and subsequently transformed into a Quadratic Unconstrained Boolean Optimization (QUBO) instance amenable to Ising and quantum-inspired solvers. We demonstrate the feasibility of this formulation on binarized MNIST by solving the resulting QUBOs with a free energy machine (FEM) solver and simulated annealing. We also show the deployment of this framework on quantum annealing and digital annealing platforms. Our results highlight the potential of quantum-inspired computing and Ising computing as a pathway toward trustworthy AI systems.

arXiv:2602.13536•Feb 14, 2026•1 citation

Exact Sampling from Neural Networks: Depth, Attention, and Decoding

Samuel Mausberg

One random output can require much less work than numerical evaluation. We study exact neural sampling, counting input probes, random bits, arithmetic, and every precision refinement. For dense tanh networks with absolute row sums at most one, weight preprocessing permits polylogarithmic expected sampling work at logarithmic depth. With zero biases, the sharp query scale is the square root of depth at every fixed first-layer rank, and also at arbitrary rank when a column envelope is bounded. The unrestricted gap between square-root and linear query cost remains open. Prefill changes attention. We construct an appendable cache index and a complete exact decoder with positive RMSNorm stabilizers. Its expected work is polynomial in model size and log T when each layer has a common affine key space of fixed rank and the context has length T. The decoder's polynomial degree grows with rank. These bounds give no useful speed prediction at the large ranks of ordinary models. A completely fixed full-rank model still has a formal polylogarithmic context bound, with potentially enormous constants. The decoder maintains finite Taylor moments, prepares precision in the background, and charges rare complete refinements by their probability. Its cost includes neural key and value creation. Without a context index, standard scaling can force linear positional search; arbitrary indexed queries face a conditional barrier in growing dimension. We also prove exact normalization factories and a matching law for an explicit additive-residual family. A checkpoint section gives the certificates and cost formulas needed to apply each positive result.

•Oct 8, 2026•0 citations

A reduction scheme for general-order Ising-like Hamiltonians in quantum heuristic solvers

Chengsi Mao, Pavel Mosharev, Yao Wang +1 more

The Ising model is ubiquitous in various optimization problems but notoriously difficult to solve due to combinatorial explosion. In view of this, Hamiltonian reduction is a useful preprocessing technique for reducing the effective problem size before applying heuristic solvers. However, existing reduction techniques mainly target second-order Ising models, whereas many pseudo-Boolean formulations naturally contain higher-order interactions. In this work, we generalize the concept of non-separable groups to arbitrary-order Ising-like models and develop a Hamiltonian reduction framework that iteratively detects and merges constrained spin groups into single variables. We benchmark the reduction on synthetic hypergraphs and higher-order network datasets, and evaluate its integration with downstream order-reduction and solver workflows. Our results establish a foundation for Hamiltonian reduction in higher-order Ising-like optimization problems.

quant-pharXiv:2607.19871•Quantum Information Processing•Jul 23, 2026•0 citations

Certified Lipschitz Bounds for Hybrid Quantum-Classical Decision Models

Sajjad Hashemian, Mohammad Saeed Arvenaghi

We study certified upper bounds on Lipschitz constants for hybrid quantum-classical decision models, which combine classical neural-network layers with parameterized quantum circuits. We pair an SDP-based certificate for classical layers with an exact spectral characterization for probability-output quantum models. For hybrid architectures, these componentwise estimates are connected through the encoding maps to obtain a certified composition bound for the complete model. The resulting bounds support sensitivity analyses relevant to robustness, stability, generalization, and fairness.

arXiv:2503.07992•Mar 11, 2025•0 citations

Noise-shaped Sampling and Annealing for Scalable On-the-edge Training of Binary and Ternary AI Models

Zihao Chen, Zhili Xiao, Ruijie Zhu +3 more

The training loss of binary and ternary neural networks, expanded as a multilinear polynomial over the discrete weights, is structurally equivalent to the energy of a hierarchical high-order Ising model. We exploit this equivalence to apply Ising solvers to discrete-weight network training, while accounting for a fundamental difference in objectives: an Ising solver seeks the ground state of a fixed energy function, whereas training seeks solutions that generalize to unseen data. Our key contribution is $ΣΔ$-Sampler, an optimizer that reconciles these objectives by embedding a stochastic sampling-and-annealing step as a noisy quantizer inside a first-order $ΣΔ$ noise-shaping loop. The loop pushes noise from weight quantization, high-order frustration, and mini-batch sampling out of the low-frequency adaptation band, so that discrete weight updates follow the smooth descent trajectories embedded in the rugged loss landscape. We validate the framework on a planted quadratic unconstrained binary optimization (QUBO) benchmark with a known minimizer; on CIFAR-10, where a $13$M-parameter ternary model comes within $1\%$ of full-precision accuracy; on ImageNet-1k, where a ternary ResNet-18 comes within $0.4\%$ Top-5 accuracy of a ternary baseline; on class-incremental continual learning, where prior-task accuracy is preserved; and on MatMul-free language models up to $2.7$B parameters, where it outperforms a straight-through-estimator baseline under matched continued pre-training. Compared with standard quantization-aware training, $ΣΔ$-Sampler reduces the training memory footprint by over $50\%$, delivers an order-of-magnitude speed-up, and reduces weight-update memory traffic by four orders of magnitude, making it attractive for on-device training and fine-tuning of large binary and ternary models.

arXiv:2610.22621•Oct 15, 2026•0 citations
We present an Ising-compatible framework for formal neural-network robustness verification under bounded input perturbations. For piecewise-linear activations, the Exact Logarithmic PWL Model (Log-PWL) provides an exact, sound, and complete formulation with a state-optimal logarithmic encoding, reducing the binary variables per neuron from linear to information-theoretically minimal logarithmic complexity. For general bounded element-wise activations, the Asymptotic Step-Envelope Model (Step-Env) uses sound piecewise-constant envelopes whose lower and upper neuron states remain decision variables coupled to a common adversarial input. We prove that its globally optimized output bounds converge uniformly to the true network extrema as the segment width vanishes, yielding asymptotic completeness of verification. We further develop a hybrid Benders solver with output-sensitive iteration bounds, leveraging a Helly-based conflict compression theory to avoid generic exponential cut explosion. Interval pruning, certificate transfer for pruned networks, and layerwise classical--Ising partitioning further reduce spin requirements. Experiments show exact certification fidelity for piecewise-linear networks and near-reference accuracy for sigmoid networks with compact spin budgets.
In recent years, quantum computers and algorithms have made significant progress indicating the prospective importance of quantum computing (QC). Especially combinatorial optimization has gained a lot of attention as an application field for near-term quantum computers, both by using gate-based QC via the Quantum Approximate Optimization Algorithm and by quantum annealing using the Ising model. However, demonstrating an advantage over classical methods in real-world applications remains an active area of research. In this work, we investigate the robustness verification of ReLU networks, which involves solving many-variable mixed-integer programs (MIPs), as a practical application. Classically, complete verification techniques struggle with large networks as the combinatorial space grows exponentially, implying that realistic networks are difficult to be verified by classical methods. To alleviate this issue, we propose to use QC for neural network verification and introduce a hybrid quantum procedure to compute provable certificates. By applying Benders decomposition, we split the MIP into a quadratic unconstrained binary optimization and a linear program which are solved by quantum and classical computers, respectively. We further improve existing hybrid methods based on the Benders decomposition by reducing the overall number of iterations and placing a limit on the maximum number of qubits required. We show that, in a simulated environment, our certificate is sound, and provides bounds on the minimum number of qubits necessary to approximate the problem. Finally, we evaluate our method within simulations and on quantum hardware.
Binary neural networks (BNNs) are increasingly deployed in edge computing applications due to their low hardware complexity and high energy efficiency. However, verifying the robustness of BNNs against input perturbations, including adversarial attacks, remains computationally challenging because the underlying decision problem is inherently combinatorial. In this paper, we propose an Ising- and quantum-inspired framework for BNN robustness verification. We show that, for a broad class of BNN architectures, robustness verification can be formulated as a Quadratic Constrained Boolean Optimization (QCBO) problem and subsequently transformed into a Quadratic Unconstrained Boolean Optimization (QUBO) instance amenable to Ising and quantum-inspired solvers. We demonstrate the feasibility of this formulation on binarized MNIST by solving the resulting QUBOs with a free energy machine (FEM) solver and simulated annealing. We also show the deployment of this framework on quantum annealing and digital annealing platforms. Our results highlight the potential of quantum-inspired computing and Ising computing as a pathway toward trustworthy AI systems.
One random output can require much less work than numerical evaluation. We study exact neural sampling, counting input probes, random bits, arithmetic, and every precision refinement. For dense tanh networks with absolute row sums at most one, weight preprocessing permits polylogarithmic expected sampling work at logarithmic depth. With zero biases, the sharp query scale is the square root of depth at every fixed first-layer rank, and also at arbitrary rank when a column envelope is bounded. The unrestricted gap between square-root and linear query cost remains open. Prefill changes attention. We construct an appendable cache index and a complete exact decoder with positive RMSNorm stabilizers. Its expected work is polynomial in model size and log T when each layer has a common affine key space of fixed rank and the context has length T. The decoder's polynomial degree grows with rank. These bounds give no useful speed prediction at the large ranks of ordinary models. A completely fixed full-rank model still has a formal polylogarithmic context bound, with potentially enormous constants. The decoder maintains finite Taylor moments, prepares precision in the background, and charges rare complete refinements by their probability. Its cost includes neural key and value creation. Without a context index, standard scaling can force linear positional search; arbitrary indexed queries face a conditional barrier in growing dimension. We also prove exact normalization factories and a matching law for an explicit additive-residual family. A checkpoint section gives the certificates and cost formulas needed to apply each positive result.
The Ising model is ubiquitous in various optimization problems but notoriously difficult to solve due to combinatorial explosion. In view of this, Hamiltonian reduction is a useful preprocessing technique for reducing the effective problem size before applying heuristic solvers. However, existing reduction techniques mainly target second-order Ising models, whereas many pseudo-Boolean formulations naturally contain higher-order interactions. In this work, we generalize the concept of non-separable groups to arbitrary-order Ising-like models and develop a Hamiltonian reduction framework that iteratively detects and merges constrained spin groups into single variables. We benchmark the reduction on synthetic hypergraphs and higher-order network datasets, and evaluate its integration with downstream order-reduction and solver workflows. Our results establish a foundation for Hamiltonian reduction in higher-order Ising-like optimization problems.
We study certified upper bounds on Lipschitz constants for hybrid quantum-classical decision models, which combine classical neural-network layers with parameterized quantum circuits. We pair an SDP-based certificate for classical layers with an exact spectral characterization for probability-output quantum models. For hybrid architectures, these componentwise estimates are connected through the encoding maps to obtain a certified composition bound for the complete model. The resulting bounds support sensitivity analyses relevant to robustness, stability, generalization, and fairness.
The training loss of binary and ternary neural networks, expanded as a multilinear polynomial over the discrete weights, is structurally equivalent to the energy of a hierarchical high-order Ising model. We exploit this equivalence to apply Ising solvers to discrete-weight network training, while accounting for a fundamental difference in objectives: an Ising solver seeks the ground state of a fixed energy function, whereas training seeks solutions that generalize to unseen data. Our key contribution is ΣΔΣΔΣΔ-Sampler, an optimizer that reconciles these objectives by embedding a stochastic sampling-and-annealing step as a noisy quantizer inside a first-order ΣΔΣΔΣΔ noise-shaping loop. The loop pushes noise from weight quantization, high-order frustration, and mini-batch sampling out of the low-frequency adaptation band, so that discrete weight updates follow the smooth descent trajectories embedded in the rugged loss landscape. We validate the framework on a planted quadratic unconstrained binary optimization (QUBO) benchmark with a known minimizer; on CIFAR-10, where a 131313M-parameter ternary model comes within 1%1\%1% of full-precision accuracy; on ImageNet-1k, where a ternary ResNet-18 comes within 0.4%0.4\%0.4% Top-5 accuracy of a ternary baseline; on class-incremental continual learning, where prior-task accuracy is preserved; and on MatMul-free language models up to 2.72.72.7B parameters, where it outperforms a straight-through-estimator baseline under matched continued pre-training. Compared with standard quantization-aware training, ΣΔΣΔΣΔ-Sampler reduces the training memory footprint by over 50%50\%50%, delivers an order-of-magnitude speed-up, and reduces weight-update memory traffic by four orders of magnitude, making it attractive for on-device training and fine-tuning of large binary and ternary models.