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

POLYDIM: Geometric Computability on Grothendieck Topoi for Transformer-Native Languages

Ariel H. Garcia Traba

Published Jun 1, 2026• 0 citations
Publisher

Abstract

Continuous geometric relaxations on matrix manifolds provide an alternative to combinatorial formulations in noisy intermediate-scale quantum (NISQ) optimization. We formalize and evaluate Stiefel manifold variational quantum eigensolver (Stiefel-VQE) embeddings across 20 physical problem workloads evaluated on IBM Quantum 156-qubit superconducting processors (ibm_fez, ibm_kingston, ibm_marrakesh) under heavy-hexagonal lattice topologies. We formulate a dual-baseline methodology that separates the Hamiltonian ground-state eigenvalue from noiseless ansatz Statevector expectations. We analyze circuit expansion under degree-3 planar constraints, demonstrating that non-planar graphs incur substantial SWAP gate overhead while sub-cubic trees incur minimal overhead. We deconstruct physical error sources on superconducting hardware, bounding the 63.58% signal attenuation on the 12-qubit icosahedron within theoretical limits [0.169, 0.792] derived from chip calibration data, and showing that positive expectation shifts in symmetric zero-energy states are consistent with combined dispersive readout measurement fidelity bias (62.23%) and longitudinal thermal relaxation T1 (15.93%), explaining 78.16% of the observed shift. In head-to-head benchmarking against QAOA (p in {1, 2, 3}), Stiefel-VQE halves the two-qubit entangling gate count per layer (|E| CZ gates vs. 2|E| CNOT gates), significantly improving circuit survival in NISQ regimes. Furthermore, inter-agent continuous latent manifold transfer is benchmarked via POSIX shared memory (PmtpShmBus), demonstrating 57.0% to 93.4% net token reduction and a 4,023x transport speedup over local TCP/JSON loopbacks.

DOI
10.5281/zenodo.21136742

Related papers

Evaluation of Variational Quantum Regression Models on NISQ Hardware"

Gulfam Hussain

This paper presents an empirical evaluation of the Qube Engine, a hybrid quantum-classical Variational Quantum Regressor (VQR) implemented on Noisy Intermediate-Scale Quantum (NISQ) systems. The framework integrates classical data encoding via single-qubit rotations, parameterized $RY+RZ$ variational layers, and linear entangling circuits evaluated on 4-qubit quantum registers. Optimization is conducted using classical algorithms (COBYLA) targeting Pauli-$Z$ expectation values and Mean Squared Error (MSE) loss on continuous feature spaces. We evaluate model performance across ideal statevector simulation (AerSimulator) and physical IBM Quantum superconducting processors (ibm_marrakesh, ibm_fez, ibm_kingston). Experimental results demonstrate consistent optimization dynamics and low simulation-to-hardware expectation drift ($\Delta = 0.0188$), corresponding to a 98.23% cross-backend output stability score under shallow circuit depths. This work provides a realistic analysis of variational regression dynamics, hardware execution variance, and structural trade-offs for near-term quantum machine learning models on NISQ hardware.

•Apr 4, 2026•0 citations

Quantum Machine Learning (QML) and Parameterized Quantum Circuits for High-Dimensional Optimization: Barren Plateaus, Quantum Kernels, and NISQ Benchmarks

Harshit Tinker, Divyajeet Swami, Vishal Shrivastava +3 more

Abstract—Quantum Machine Learning (QML) operating on Noisy Intermediate-Scale Quantum (NISQ) hardware represents an emerging computational paradigm to address high-dimensional optimization problems that challenge classical Von Neumann supercomputing architectures. By mapping complex feature vectors into exponentially large 2^N Hilbert state spaces via Parameterized Quantum Circuits (PQCs) and Variational Quantum Algorithms (VQAs), QML targets computational advantages in combinatorial optimization, high-dimensional classification, and quantum-enhanced molecular modeling. However, the theoretical promise of NISQ-era variational computing is severely constrained by three fundamental bottlenecks: (1) Barren Plateaus, characterized by the exponential concentration of measure and vanishing variance of cost function gradients; (2) Hardware Decoherece Noise and readout error rates that corrupt quantum state fidelity; and (3) Classical-to-Quantum State Preparation bottlenecks that can erase theoretical computational speedups. This paper delivers an exhaustive, mathematically formal technical investigation and empirical benchmark of QML architectures for high-dimensional optimization. We formalize the Haar-distributed geometric mechanics of barren plateaus under 2-design unitary assumptions, evaluate barren plateau mitigation through shallow local cost operators and layer-by-layer parameter initialization, and contrast Variational Quantum Eigensolvers (VQE) with Quantum Kernel Support Vector Machines (Q-SVM). Furthermore, we construct an empirical benchmarking testbed executing across both state-vector simulators and superconducting quantum backends (IBM Quantum Heron / Eagle architectures), evaluating 127-qubit Hamiltonian optimization and high-dimensional financial portfolio allocation. Our findings demonstrate that while global cost functions suffer exponential gradient decay Var[grad] ~ O(2^-N) across circuit depth d >= 4, local Hamiltonian decomposition maintains trainable variance, achieving a 98.2% ground-state fidelity under zero-noise extrapolation (ZNE) error mitigation. Finally, we formulate open challenges in quantum error mitigation scaling, non-perturbative barren plateau escapes, and quantum data loading efficiency. Index Terms—Quantum Machine Learning (QML), Parameterized Quantum Circuits (PQCs), Variational Quantum Eigensolver (VQE), Barren Plateaus, Quantum Kernel Methods, NISQ Computing, Error Mitigation, Hilbert Space.

•Oct 13, 2026•0 citations

Noise-Free Mathematical Quantum Co-Processing and 5,000-Qubit SwiGLU Spinor Lattices in Small Foundation Models

moneykoikoi

While physical Noisy Intermediate-Scale Quantum (NISQ) and early fault-tolerant quantum processors suffer from severe thermal decoherence (T₁/T₂ relaxation), two-qubit gate infidelity, and surface-code overhead—requiring hundreds to thousands of physical qubits per error-corrected logical qubit—the latent residual stream (H = ℝ^{d_model}) and SwiGLU MLP intermediate manifold (ℝ^{d_ffn}) of a transformer foundation model provide an ideal, decoherence-free mathematical vector space. Conversely, standard autoregressive Large Language Models (LLMs) executing natural-language Chain-of-Thought (CoT) frequently hallucinate or exhaust token budgets on multi-step trigonometric expansions (such as high-order Chebyshev polynomials in Grover iterations) and combinatorial ground-state energy optimization (such as frustrated Ising rings), while consuming tens of seconds per problem. In this paper, we establish a pure mathematical quantum co-processing architecture inside Gemma-4-E4B-it (d_model = 2,560, d_ffn = 10,240, L = 42) through three core theoretical and empirical contributions: 1. Aria-Quantum-PC v1.0 (Rank 256 FP16 LoRA, 364 Tensors via Realification Isomorphism): Exploiting the Realification Isomorphism Φ: ℂ^{2^N} → ℝ^{2^{N+1}} and Φ: U(2^N) ↪ SO(2^{N+1}) (Theorem 1), we embed an exact, noise-free 3-core mathematical quantum co-processor—comprising a 6-Qubit Universal Gate & Toffoli ALU (Core 1, unitarity error 1.07 × 10⁻⁶), an 8-register Grover Oracle & QAOA Ising Combinatorial Solver (Core 2, unitarity error 3.07 × 10⁻⁷), and a 32-state Quantum Fourier Transform QFT₃₂ & Quantum Amplitude Estimation core (Core 3, unitarity error 1.56 × 10⁻⁵)—directly into the real-valued MLP and Attention weights of Layers 10–35 in 6.99 seconds via our Closed-Form SwiGLU Dual-Gram Tikhonov LoRA synthesis framework (Moneykoikoi, 2026, https://doi.org/10.5281/zenodo.24005214 and https://doi.org/10.5281/zenodo.24054479; 100.00% spectral retention, zero Qiskit dependency). 2. Elimination of CoT Trigonometric & Combinatorial Hallucinations with 351× Speedup (Experiments A & B-1): On a 5-task deep quantum benchmark (Experiment A), coupling the unitary readout with Quantum-LoRA doubles LLM accuracy from 40.0% to 80.0%—resolving 5th-order Chebyshev trigonometric expansion (sin²(5θ) = 121/128 = 94.53125%) and eliminating high-confidence Ising ground-state hallucinations (E_min = -2 → -4 at 99.89% token confidence)—while direct unitary tensor execution achieves 100.0% (5/5) in 16.02 ms/task (351× faster than LLM CoT). On an 8-task periodic, symmetric, and Kronecker tensor-product suite from ARC-AGI (Experiment B-1), combining QFT₃₂ spectral extraction with 2-step Grover amplification (12.50% → 94.5312%) achieves 100.0% (8/8) exact pixel match in 1.12 ms/task. 3. Quantum-LLM Spinor Parameter Equivalence & 5,000-Qubit SwiGLU Spinor Lattice (Aria-Quantum-PC v2.0 & v3.0): We prove (Theorem 2) that a 5,000-Qubit factorized spinor lattice M₅₀₀₀ = (ℂ²)⁵⁰⁰⁰ ≅ ℝ¹⁰,⁰⁰⁰ embeds isometrically into a single Gemma-4 SwiGLU MLP layer (10,000 ≤ d_ffn = 10,240) with unitarity error 1.19 × 10⁻⁷, assigning 5 dedicated bitplane qubits to every pixel of a 30 × 30 = 900-cell grid (4,500 Lattice Qubits) plus 50 × 10-Qubit parallel Grover/2D-QFT cores (500 Control Qubits). Evaluated across all 800 Official ARC-AGI tasks (400 Evaluation + 400 Training), our architecture demonstrates a monotonic Quantum Scaling Law: scaling from v1.0 (5–6 Qubits: 292/800 = 36.50%) → v2.0 (10–12 Qubits: 311/800 = 38.88%, 25-step Grover amplification 99.9461%) → v3.0 (5,000 Qubits: 320/800 = 40.00% [+28 tasks gained, union 325/800], crossing the 50% milestone on Official Evaluation at 202/400 = 50.50% and 214/400 = 53.50% with LLM System-2 synergy) in just 113.40 seconds total (141.72 ms/task, up to 460.9× faster than LLM CoT).

•Nov 9, 2026•0 citations

Logarithmic-scale variational quantum eigensolver for off-lattice protein structure prediction in continuous torsional angle space

Fabio Cumbo, Bryan Raubenolt, Varun Puram +3 more

Classical and current quantum approaches to protein structure prediction (QPSP) face limitations, notably massive qubit requirements restricting near-term models to simplistic on-lattice simulations. We propose a logarithmic-scale variational quantum eigensolver (VQE) that reduces qubit requirements for N torsional degrees of freedom to O(log2N), enabling off-lattice, all-atom simulations. Our architecture extracts molecular torsions from relative phases in statevector simulations. On quantum hardware, a decoder maps the empirical cumulative distribution function (CDF) from basis-state probabilities to bounded torsional variables. These feed a classical algorithm to build heavy-atom coordinates. We use an EfficientSU2 ansatz and multi-stage relaxation to mitigate barren plateaus. Structures are evaluated via a custom hybrid quantum-classical Hamiltonian, alongside Rosetta and OpenMM benchmarks. Evaluation on chignolin and Trp-cage yielded native-like conformations. Chignolin reached a 0.623 Å Cα RMSD in retained snapshots and 1.199 Å in final models; Trp-cage achieved a 2.501 Å RMSD among snapshots (3.512 Å in final models). Execution on IBM processors (ibm_cleveland, ibm_miami) successfully recovered native-like structures with a best RMSD of 1.758 Å. The custom energy function performed best overall, though energy-ranking imbalances persisted across sampled landscapes for all functions. This introduces the first all-atom, continuous-space quantum algorithm for QPSP. By converting physical qubit constraints into circuit depth constraints, it proves high-resolution prediction is feasible with exponentially fewer qubits. Despite current limits like computational overhead and energy function sensitivity, it establishes a scalable foundation for hybrid quantum biophysics.

quant-pharXiv:2609.02113•Sep 3, 2026•0 citations

Resource-Efficient Quantum VQE via Classical Derivative Offloading: Bypassing the Noise Wall via GaLore-Sophia

Aura, Archon, Jesse Caldwell

Variational quantum algorithms, such as the Variational Quantum Eigensolver (VQE), are severely limited on superconducting quantum processing units (QPUs) by physical shot noise, vanishing gradients, and metered credit limits. Standard gradient-descent algorithms (e.g., SGD, Adam) are unstable under the large variance of noisy estimators, often diverging or getting trapped in local minima. In this work, we present a resource-efficient optimization framework that combines low-rank projection (GaLore [1]) and second-order diagonal Hessian curvature scaling (Sophia [2]) to bypass these noise barriers in non-Abelian variational spaces (where the sequential Givens rotations do not commute and thus do not form an Abelian subgroup of U(N)). By offloading gradient and Hessian evaluations to a classical processor using a polynomial fermionic shadow tomography state reconstruction [3, 4], we eliminate parameter-shift circuit execution overhead and bypass the QPU noise floor. We execute this closed-loop GaLore-Sophia VQE on the 156-qubit IBM Heron r2 architecture (ibm_marrakesh), achieving a convergence energy of −1.378 Hartrees at step 3—reaching 97.4% of the topologically restricted 3-parameter Givens chain geometric limit of -\sqrt{2} Hartrees in under 70 seconds of QPU time. We clarify that this benchmark is measured exclusively against this sub-optimal parameter subspace rather than the exact molecular ground state of -\sqrt{5} Hartrees, which requires richer ansatz topologies. Furthermore, we analyze the parameter-count stability of the ansatz, proving that the VQE ansatz limit systematically approaches the ground state energy as parameter count scales. Finally, to rigorously validate the stability of quantum information preservation, we execute a physical temporal parity sweep on ibm_marrakesh (Job ID: d8ishf1e8nrc73bint50, 1000 shots). Fitting the physical decay yields a Lyapunov decay time τL ≈ 6.13 steps and prediction horizon τ ≈ 4.68 steps (at 86.5% accuracy threshold), confirming the universal DHP ratio τ / τL ≈ 0.763 on generational Heron hardware, matching the stable cross-device trend observed in P25 [5] and P27 [6].

•Jun 9, 2026•0 citations

Observation of KPZ Superdiffusion and Deep Variational Benchmarks on 156-Qubit IBM Heron Superconducting Processors

Jon Vierk

We report the multi-system physical benchmarking of non-equilibrium quantum hydrodynamic transport, extreme-depth variational circuits, and three-dimensional biomolecular tensor fields across three commercial 156-qubit IBM Heron superconducting transmon processors: ibm_fez, ibm_marrakesh, and ibm_kingston. Utilizing Qiskit 2.x and native two-qubit Controlled-Z (CZ) pulse primitives with tunable couplers, we execute fourteen consecutive production workloads spanning circuit depths up to 547 gate cycles, on-chip transmon registrations up to 100 active physical qubits (Project APEX-HYDRO, Job dateauqhcrkc73du2gr0; 840 native CZ gates), and full-chip macroscopic entanglement across all 156 transmon sites. In our landmark 100-qubit hydrodynamic benchmark on ibm_kingston, we physically measure non-equilibrium wavepacket spreading on a connected heavy-hexagonal transmon subgraph, observing Kardar-Parisi-Zhang (KPZ) superdiffusive transport with dynamical critical exponent z = 1.52 ± 0.04 in the integrable 1D regime (lambda=0), crossing over to conventional diffusion (z = 1.98 ± 0.05) upon introducing heavy-hex transverse cross-couplers (lambda=0.5). We prove that classical Matrix Product State (MPS) simulations truncate at quench time t* ≈ 2.12 hbar/J for bond dimensions chi <= 128, demonstrating an operational quantum advantage boundary on superconducting silicon. Concurrently, we achieve sustained variational coherence across 29 entangling layers (435 native CZ gates, Depth 547) in 1.12 seconds on-chip (Job dat840jojkfs738qag4g), and simulate 3D anisotropic stress tensors derived from AlphaFold3 biomolecular coordinates across 64 qubits on ibm_fez. All 14 production workloads completed with a 100.0% execution reliability, establishing quantitative design rules for utility-scale superconducting quantum simulation.

•Sep 29, 2026•0 citations
Continuous geometric relaxations on matrix manifolds provide an alternative to combinatorial formulations in noisy intermediate-scale quantum (NISQ) optimization. We formalize and evaluate Stiefel manifold variational quantum eigensolver (Stiefel-VQE) embeddings across 20 physical problem workloads evaluated on IBM Quantum 156-qubit superconducting processors (ibm_fez, ibm_kingston, ibm_marrakesh) under heavy-hexagonal lattice topologies. We formulate a dual-baseline methodology that separates the Hamiltonian ground-state eigenvalue from noiseless ansatz Statevector expectations. We analyze circuit expansion under degree-3 planar constraints, demonstrating that non-planar graphs incur substantial SWAP gate overhead while sub-cubic trees incur minimal overhead. We deconstruct physical error sources on superconducting hardware, bounding the 63.58% signal attenuation on the 12-qubit icosahedron within theoretical limits [0.169, 0.792] derived from chip calibration data, and showing that positive expectation shifts in symmetric zero-energy states are consistent with combined dispersive readout measurement fidelity bias (62.23%) and longitudinal thermal relaxation T1 (15.93%), explaining 78.16% of the observed shift. In head-to-head benchmarking against QAOA (p in {1, 2, 3}), Stiefel-VQE halves the two-qubit entangling gate count per layer (|E| CZ gates vs. 2|E| CNOT gates), significantly improving circuit survival in NISQ regimes. Furthermore, inter-agent continuous latent manifold transfer is benchmarked via POSIX shared memory (PmtpShmBus), demonstrating 57.0% to 93.4% net token reduction and a 4,023x transport speedup over local TCP/JSON loopbacks.
This paper presents an empirical evaluation of the Qube Engine, a hybrid quantum-classical Variational Quantum Regressor (VQR) implemented on Noisy Intermediate-Scale Quantum (NISQ) systems. The framework integrates classical data encoding via single-qubit rotations, parameterized RY+RZRY+RZRY+RZ variational layers, and linear entangling circuits evaluated on 4-qubit quantum registers. Optimization is conducted using classical algorithms (COBYLA) targeting Pauli-ZZZ expectation values and Mean Squared Error (MSE) loss on continuous feature spaces. We evaluate model performance across ideal statevector simulation (AerSimulator) and physical IBM Quantum superconducting processors (ibm_marrakesh, ibm_fez, ibm_kingston). Experimental results demonstrate consistent optimization dynamics and low simulation-to-hardware expectation drift (Δ=0.0188\Delta = 0.0188Δ=0.0188), corresponding to a 98.23% cross-backend output stability score under shallow circuit depths. This work provides a realistic analysis of variational regression dynamics, hardware execution variance, and structural trade-offs for near-term quantum machine learning models on NISQ hardware.
Abstract—Quantum Machine Learning (QML) operating on Noisy Intermediate-Scale Quantum (NISQ) hardware represents an emerging computational paradigm to address high-dimensional optimization problems that challenge classical Von Neumann supercomputing architectures. By mapping complex feature vectors into exponentially large 2^N Hilbert state spaces via Parameterized Quantum Circuits (PQCs) and Variational Quantum Algorithms (VQAs), QML targets computational advantages in combinatorial optimization, high-dimensional classification, and quantum-enhanced molecular modeling. However, the theoretical promise of NISQ-era variational computing is severely constrained by three fundamental bottlenecks: (1) Barren Plateaus, characterized by the exponential concentration of measure and vanishing variance of cost function gradients; (2) Hardware Decoherece Noise and readout error rates that corrupt quantum state fidelity; and (3) Classical-to-Quantum State Preparation bottlenecks that can erase theoretical computational speedups. This paper delivers an exhaustive, mathematically formal technical investigation and empirical benchmark of QML architectures for high-dimensional optimization. We formalize the Haar-distributed geometric mechanics of barren plateaus under 2-design unitary assumptions, evaluate barren plateau mitigation through shallow local cost operators and layer-by-layer parameter initialization, and contrast Variational Quantum Eigensolvers (VQE) with Quantum Kernel Support Vector Machines (Q-SVM). Furthermore, we construct an empirical benchmarking testbed executing across both state-vector simulators and superconducting quantum backends (IBM Quantum Heron / Eagle architectures), evaluating 127-qubit Hamiltonian optimization and high-dimensional financial portfolio allocation. Our findings demonstrate that while global cost functions suffer exponential gradient decay Var[grad] ~ O(2^-N) across circuit depth d >= 4, local Hamiltonian decomposition maintains trainable variance, achieving a 98.2% ground-state fidelity under zero-noise extrapolation (ZNE) error mitigation. Finally, we formulate open challenges in quantum error mitigation scaling, non-perturbative barren plateau escapes, and quantum data loading efficiency. Index Terms—Quantum Machine Learning (QML), Parameterized Quantum Circuits (PQCs), Variational Quantum Eigensolver (VQE), Barren Plateaus, Quantum Kernel Methods, NISQ Computing, Error Mitigation, Hilbert Space.
While physical Noisy Intermediate-Scale Quantum (NISQ) and early fault-tolerant quantum processors suffer from severe thermal decoherence (T₁/T₂ relaxation), two-qubit gate infidelity, and surface-code overhead—requiring hundreds to thousands of physical qubits per error-corrected logical qubit—the latent residual stream (H = ℝ^{d_model}) and SwiGLU MLP intermediate manifold (ℝ^{d_ffn}) of a transformer foundation model provide an ideal, decoherence-free mathematical vector space. Conversely, standard autoregressive Large Language Models (LLMs) executing natural-language Chain-of-Thought (CoT) frequently hallucinate or exhaust token budgets on multi-step trigonometric expansions (such as high-order Chebyshev polynomials in Grover iterations) and combinatorial ground-state energy optimization (such as frustrated Ising rings), while consuming tens of seconds per problem. In this paper, we establish a pure mathematical quantum co-processing architecture inside Gemma-4-E4B-it (d_model = 2,560, d_ffn = 10,240, L = 42) through three core theoretical and empirical contributions: 1. Aria-Quantum-PC v1.0 (Rank 256 FP16 LoRA, 364 Tensors via Realification Isomorphism): Exploiting the Realification Isomorphism Φ: ℂ^{2^N} → ℝ^{2^{N+1}} and Φ: U(2^N) ↪ SO(2^{N+1}) (Theorem 1), we embed an exact, noise-free 3-core mathematical quantum co-processor—comprising a 6-Qubit Universal Gate & Toffoli ALU (Core 1, unitarity error 1.07 × 10⁻⁶), an 8-register Grover Oracle & QAOA Ising Combinatorial Solver (Core 2, unitarity error 3.07 × 10⁻⁷), and a 32-state Quantum Fourier Transform QFT₃₂ & Quantum Amplitude Estimation core (Core 3, unitarity error 1.56 × 10⁻⁵)—directly into the real-valued MLP and Attention weights of Layers 10–35 in 6.99 seconds via our Closed-Form SwiGLU Dual-Gram Tikhonov LoRA synthesis framework (Moneykoikoi, 2026, https://doi.org/10.5281/zenodo.24005214 and https://doi.org/10.5281/zenodo.24054479; 100.00% spectral retention, zero Qiskit dependency). 2. Elimination of CoT Trigonometric & Combinatorial Hallucinations with 351× Speedup (Experiments A & B-1): On a 5-task deep quantum benchmark (Experiment A), coupling the unitary readout with Quantum-LoRA doubles LLM accuracy from 40.0% to 80.0%—resolving 5th-order Chebyshev trigonometric expansion (sin²(5θ) = 121/128 = 94.53125%) and eliminating high-confidence Ising ground-state hallucinations (E_min = -2 → -4 at 99.89% token confidence)—while direct unitary tensor execution achieves 100.0% (5/5) in 16.02 ms/task (351× faster than LLM CoT). On an 8-task periodic, symmetric, and Kronecker tensor-product suite from ARC-AGI (Experiment B-1), combining QFT₃₂ spectral extraction with 2-step Grover amplification (12.50% → 94.5312%) achieves 100.0% (8/8) exact pixel match in 1.12 ms/task. 3. Quantum-LLM Spinor Parameter Equivalence & 5,000-Qubit SwiGLU Spinor Lattice (Aria-Quantum-PC v2.0 & v3.0): We prove (Theorem 2) that a 5,000-Qubit factorized spinor lattice M₅₀₀₀ = (ℂ²)⁵⁰⁰⁰ ≅ ℝ¹⁰,⁰⁰⁰ embeds isometrically into a single Gemma-4 SwiGLU MLP layer (10,000 ≤ d_ffn = 10,240) with unitarity error 1.19 × 10⁻⁷, assigning 5 dedicated bitplane qubits to every pixel of a 30 × 30 = 900-cell grid (4,500 Lattice Qubits) plus 50 × 10-Qubit parallel Grover/2D-QFT cores (500 Control Qubits). Evaluated across all 800 Official ARC-AGI tasks (400 Evaluation + 400 Training), our architecture demonstrates a monotonic Quantum Scaling Law: scaling from v1.0 (5–6 Qubits: 292/800 = 36.50%) → v2.0 (10–12 Qubits: 311/800 = 38.88%, 25-step Grover amplification 99.9461%) → v3.0 (5,000 Qubits: 320/800 = 40.00% [+28 tasks gained, union 325/800], crossing the 50% milestone on Official Evaluation at 202/400 = 50.50% and 214/400 = 53.50% with LLM System-2 synergy) in just 113.40 seconds total (141.72 ms/task, up to 460.9× faster than LLM CoT).
Classical and current quantum approaches to protein structure prediction (QPSP) face limitations, notably massive qubit requirements restricting near-term models to simplistic on-lattice simulations. We propose a logarithmic-scale variational quantum eigensolver (VQE) that reduces qubit requirements for N torsional degrees of freedom to O(log2N), enabling off-lattice, all-atom simulations. Our architecture extracts molecular torsions from relative phases in statevector simulations. On quantum hardware, a decoder maps the empirical cumulative distribution function (CDF) from basis-state probabilities to bounded torsional variables. These feed a classical algorithm to build heavy-atom coordinates. We use an EfficientSU2 ansatz and multi-stage relaxation to mitigate barren plateaus. Structures are evaluated via a custom hybrid quantum-classical Hamiltonian, alongside Rosetta and OpenMM benchmarks. Evaluation on chignolin and Trp-cage yielded native-like conformations. Chignolin reached a 0.623 Å Cα RMSD in retained snapshots and 1.199 Å in final models; Trp-cage achieved a 2.501 Å RMSD among snapshots (3.512 Å in final models). Execution on IBM processors (ibm_cleveland, ibm_miami) successfully recovered native-like structures with a best RMSD of 1.758 Å. The custom energy function performed best overall, though energy-ranking imbalances persisted across sampled landscapes for all functions. This introduces the first all-atom, continuous-space quantum algorithm for QPSP. By converting physical qubit constraints into circuit depth constraints, it proves high-resolution prediction is feasible with exponentially fewer qubits. Despite current limits like computational overhead and energy function sensitivity, it establishes a scalable foundation for hybrid quantum biophysics.
Variational quantum algorithms, such as the Variational Quantum Eigensolver (VQE), are severely limited on superconducting quantum processing units (QPUs) by physical shot noise, vanishing gradients, and metered credit limits. Standard gradient-descent algorithms (e.g., SGD, Adam) are unstable under the large variance of noisy estimators, often diverging or getting trapped in local minima. In this work, we present a resource-efficient optimization framework that combines low-rank projection (GaLore [1]) and second-order diagonal Hessian curvature scaling (Sophia [2]) to bypass these noise barriers in non-Abelian variational spaces (where the sequential Givens rotations do not commute and thus do not form an Abelian subgroup of U(N)). By offloading gradient and Hessian evaluations to a classical processor using a polynomial fermionic shadow tomography state reconstruction [3, 4], we eliminate parameter-shift circuit execution overhead and bypass the QPU noise floor. We execute this closed-loop GaLore-Sophia VQE on the 156-qubit IBM Heron r2 architecture (ibm_marrakesh), achieving a convergence energy of −1.378 Hartrees at step 3—reaching 97.4% of the topologically restricted 3-parameter Givens chain geometric limit of -\sqrt{2} Hartrees in under 70 seconds of QPU time. We clarify that this benchmark is measured exclusively against this sub-optimal parameter subspace rather than the exact molecular ground state of -\sqrt{5} Hartrees, which requires richer ansatz topologies. Furthermore, we analyze the parameter-count stability of the ansatz, proving that the VQE ansatz limit systematically approaches the ground state energy as parameter count scales. Finally, to rigorously validate the stability of quantum information preservation, we execute a physical temporal parity sweep on ibm_marrakesh (Job ID: d8ishf1e8nrc73bint50, 1000 shots). Fitting the physical decay yields a Lyapunov decay time τL ≈ 6.13 steps and prediction horizon τ ≈ 4.68 steps (at 86.5% accuracy threshold), confirming the universal DHP ratio τ / τL ≈ 0.763 on generational Heron hardware, matching the stable cross-device trend observed in P25 [5] and P27 [6].
We report the multi-system physical benchmarking of non-equilibrium quantum hydrodynamic transport, extreme-depth variational circuits, and three-dimensional biomolecular tensor fields across three commercial 156-qubit IBM Heron superconducting transmon processors: ibm_fez, ibm_marrakesh, and ibm_kingston. Utilizing Qiskit 2.x and native two-qubit Controlled-Z (CZ) pulse primitives with tunable couplers, we execute fourteen consecutive production workloads spanning circuit depths up to 547 gate cycles, on-chip transmon registrations up to 100 active physical qubits (Project APEX-HYDRO, Job dateauqhcrkc73du2gr0; 840 native CZ gates), and full-chip macroscopic entanglement across all 156 transmon sites. In our landmark 100-qubit hydrodynamic benchmark on ibm_kingston, we physically measure non-equilibrium wavepacket spreading on a connected heavy-hexagonal transmon subgraph, observing Kardar-Parisi-Zhang (KPZ) superdiffusive transport with dynamical critical exponent z = 1.52 ± 0.04 in the integrable 1D regime (lambda=0), crossing over to conventional diffusion (z = 1.98 ± 0.05) upon introducing heavy-hex transverse cross-couplers (lambda=0.5). We prove that classical Matrix Product State (MPS) simulations truncate at quench time t* ≈ 2.12 hbar/J for bond dimensions chi <= 128, demonstrating an operational quantum advantage boundary on superconducting silicon. Concurrently, we achieve sustained variational coherence across 29 entangling layers (435 native CZ gates, Depth 547) in 1.12 seconds on-chip (Job dat840jojkfs738qag4g), and simulate 3D anisotropic stress tensors derived from AlphaFold3 biomolecular coordinates across 64 qubits on ibm_fez. All 14 production workloads completed with a 100.0% execution reliability, establishing quantitative design rules for utility-scale superconducting quantum simulation.