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

Replication Package and Codebase: A Tight Lower Bound on Generalization in Quantum Machine Learning via Dynamical Lie Algebra Dimension

Ayad Alhusseiny

Published Oct 16, 2026• 0 citations
Publisher

Abstract

Orbit dimension, not parameter count, bounds quantum sample complexity. For continuous parameterized quantum circuits on $n$ qubits, the minimax estimation risk obeys \[ \mathcal{R}(d_{\mathrm{eff}}, M) \;\geq\; \Omega\!\left(\sqrt{\frac{d_{\mathrm{eff}}}{M}}\right), \] where $d_{\mathrm{eff}} = \dim(\mathcal{O}_O)$ is the real dimension of the observable's adjoint orbit under the dynamical Lie group. For balanced Pauli observables $O = Z^{\otimes n}$, quotienting the dynamical Lie algebra by the stabilizer subgroup yields the complex Grassmannian $Gr(2^{n-1}, 2^n)$ of real dimension $d_{\mathrm{eff}} = 2^{2n-1} = \frac{1}{2}4^n \approx \frac{1}{2}\dim(\mathfrak{g})$, which matches upper bounds of Caro et al.~(2022) and Ohno~(2025) up to a universal constant and resolves the exact tightness rate $\Theta(\sqrt{\dim(\mathfrak{g})/M})$. For rank-1 projective observables, the orbit contracts instead to complex projective space $\mathbb{CP}^{2^n-1}$ with dimension $d_{\mathrm{eff}} = 2(2^n-1)$. Sample requirements therefore explode exponentially with qubit count. The scaling identifies Wang \& Wu's (2025) dimension-independent $\mathcal{O}(1/\sqrt{M})$ uniform convergence bound as an artifact of uniform stability rather than true estimation power. PennyLane benchmarks across $n=1,2,3,4$ qubits confirm this exponential threshold. Barren plateaus and statistical sample bottlenecks stem from an identical geometric origin: volume growth in $SU(2^n)$ simultaneously suppresses gradient variances as $\sim 1/\dim(\mathfrak{g})$ and drives sample complexity up to $\sim \sqrt{\dim(\mathfrak{g})/M}$.

DOI
10.5281/zenodo.23410873

Related papers

A Fundamental Duality in Quantum Machine Learning: Trainability vs. Generalization via Dynamical Lie Algebras

Ayad Alhusseiny

Quantum learnability is an algebraic invariant. While optimization and statistical generalization have long been analyzed through disconnected mathematical formalisms in quantum machine learning, both phenomena emerge as dual geometric projections of the dynamical Lie algebra (DLA) dimension g = Lie{-i H_k} ⊆ su(2^n); across any parameterized circuit family, polynomial trainability (Var[∇C] ≥ Ω(1/poly(n))) and polynomial generalizability (M*(ε) ≤ O(poly(n)/ε^2)) are strictly equivalent and hold if and only if dim(g_n) ∈ O(poly(n)). For any observable with adjoint orbit O_O, the operational product satisfies Var[∇C] · M*(ε) = Θ(dim(O_O) / (dim(g_n) ε^2)), which collapses to the universal, dimension-free invariant Θ(1/ε^2) for all generic, non-degenerate observables, identically independent of system size n and algebraic dimension dim(g_n). Unconstrained expressivity (dim(g_n) ~ 4^n) triggers a simultaneous catastrophe: exponential gradient vanishing accompanied by exponential sample complexity explosion. Subalgebra confinement cures both failures. By restricting generators to permutation-equivariant or symmetry-protected subalgebras, the circuit enforces dim(g_n) ≤ O(poly(n)) and secures non-vanishing gradient flow alongside statistical sample efficiency. Exact statevector simulations in PennyLane across n ∈ {2, 3, 4, 5, 6} corroborate this duality: hardware-efficient ansätze exhibit a 17.22× variance collapse, whereas symmetry-protected models sustain Θ(1) gradients and achieve twofold lower test error.

•Oct 23, 2026•0 citations

Scaling Behavior of Parameterized Quantum Circuits from a Lie-Algebraic Perspective

Hiroshi Ohno

Understanding how the performance of parameterized quantum circuits scales with available resources is important for characterizing their trainability and effective model capacity. In this study, we numerically investigate data scaling, model scaling, and compute scaling in parameterized quantum circuits and examine Lie-algebraic quantities as alternative measures of model size. In addition to the number of circuit parameters, we consider the dimension of the dynamical Lie algebra, the observable-orbit dimension, and a Jacobian effective dimension defined as the rank of the Jacobian of the parameterized observable orbit. Using a regression task with randomly generated Pauli-string generators, we observe decreasing loss with increasing training dataset size, parameter size, and number of optimization iterations over the ranges investigated. For model scaling, the dynamical Lie algebra and observable orbit dimensions rapidly saturate as the parameter size increases, whereas the Jacobian effective dimension remains strongly correlated with the parameter size and exhibits comparable scaling behavior. These results suggest that the Jacobian effective dimension provides a geometry-aware measure of the locally accessible observable degrees of freedom of finite-depth parameterized quantum circuits and may serve as a useful scaling parameter beyond the nominal parameter size.

quant-pharXiv:2608.20603•Aug 24, 2026•0 citations

Generalization in Data Re-uploading Quantum Neural Networks: A Dynamical Lie Algebraic and Harmonic Analysis Perspective

Ayad Alhusseiny

Commutation failure controls quantum generalization. When data encoding gates fail to commute with trainable parameter gates across consecutive re-uploading layers, the effective dynamical Lie algebra becomes data-dependent; this dependency invalidates the parameter-isolation assumption underpinning existing Lie-algebraic frameworks (Caro et al., 2022; Ohno, 2025). We resolve this structural gap by proving that the statistical generalization capacity of L-layer data re-uploading circuits is tightly bounded by the effective hypothesis dimension: d_eff(L) = min(L * p_layer, |Ω_L|) <= min(L * p_layer, (2L + 1)^d), where d is the input data dimension, p_layer the number of trainable parameters per layer, and |Ω_L| the cardinality of accessible Fourier frequencies generated through non-commuting Lie brackets in the joint dynamical Lie algebra g_joint = Lie_R{i H_j, i G_k}. Harmonic analysis combined with metric entropy chaining establishes matching upper and lower bounds; together, they settle the exact minimax estimation rate R*(F_L, M) = \tilde{\Theta}(\sqrt{d_eff(L)/M}). Two distinct orbit concepts govern learning. The point-wise adjoint orbit dimension dim(O_O) <= dim(g_joint) at fixed x quantifies single-point state variation, whereas the global function-space Fourier capacity |Ω_L| governs sample complexity across the full data domain; conflating these quantities yields vacuous bounds. An initial basis rotation unlocks harmonic modes up to k_max = L; without this rotation, eigenstate initialization constrains the cutoff to k_max = L - 1. Bandwidth dictates depth. For target functions with harmonic bandlimit k_target, the network underfits when L < k_target and achieves full expressivity convergence once L >= k_target. Permutation-equivariant data encoding compresses the joint DLA to an su(2) subalgebra of dimension 3. This algebraic confinement suppresses the 4^n - 1 dimensional explosion across arbitrary qubit counts and shields the circuit from the curse of dimensionality. Numerical benchmarks in PennyLane across L in {1, 2, 4, 8} and sample sizes M in [10, 200] corroborate this harmonic-algebraic scaling and its associated uniform convergence rates.

•Oct 21, 2026•0 citations

Scalable Quantum Machine Learning: Trainability, Expressivity and Efficiency

Iordanis Kerenidis

Scalable parameterized quantum circuits for machine learning face three obstacles: barren plateaus, no guarantee the learned function class is classically hard, and prohibitive circuit evaluations per gradient step. We propose the unitary brick-wall: a $k$-particle fermionic architecture for nearest-neighbor hardware, combining Reconfigurable Beam Splitters with interleaved single-qubit phases and a non-Gaussian magic-state encoding, where $k$ trades classical hardness against training cost. Trainable. The brick-wall has dynamical Lie algebra $\mathfrak{u}(n)$ and is surjective onto $U(n)$ via Givens rotations, enabling Haar initialization. Two-body correlator readouts achieve gradient variance $Θ(k^3/n^5)$, polynomial in $n$ throughout $n-2k=Ω(n)$. Expressive. Hardness is controlled by $k$: best-known classical sampling runs in time $2^{Θ(k)}$, worst-case #P-hardness holds from $k=n^ε$, and the average-case machinery of Fermion Sampling applies at $k=Θ(n)$. The sampler is input-dependent, so costs are paid per query: at our operating point $k=100$ one query needs $\sim 10^{15}$ operations, beyond commercial query rates, and at $k=120$ an exascale machine sustains under one query per second. Efficient. A multi-layer parallel parameter-shift rule computes all $O(n^2)$ gradients from $4kn$ circuit evaluations per step, a factor $n/k$ reduction over the standard rule's $4n^2$, growing linearly with $n$ at fixed $k$. The unitary butterfly targets all-to-all hardware, with $2\log n$ layers of parallel gates and $(3/2)n\log n$ parameters, similar hardness guarantees, and $4k\log n$ evaluations per step -- the same factor-$n/k$ reduction. Its trainability holds at two levels: absence of exponential barren plateaus is unconditional, while the sharp $Θ(k^3/n^5)$ rate needs a conjectured scalar comparison with the two-particle Haar second moment.

quant-pharXiv:2607.24014•Jul 28, 2026•7 citations

LieTrunc-QNN: Lie Algebra Truncation and Quantum Expressivity Phase Transition from LiePrune to Provably Stable Quantum Neural Networks

Haijian Shao, Dalong A. Zhao, Xing Deng +2 more

Quantum Machine Learning (QML) is fundamentally limited by two challenges: barren plateaus (exponentially vanishing gradients) and the fragility of parameterized quantum circuits under noise. Despite extensive empirical studies, a unified theoretical framework remains lacking. We introduce LieTrunc-QNN, an algebraic-geometric framework that characterizes trainability via Lie-generated dynamics. Parameterized quantum circuits are modeled as Lie subalgebras of u(2^n), whose action induces a Riemannian manifold of reachable quantum states. Expressivity is reinterpreted as intrinsic manifold dimension and geometry. We establish a geometric capacity-plateau principle: increasing effective dimension leads to exponential gradient suppression due to concentration of measure. By restricting to structured Lie subalgebras (LieTrunc), the manifold is contracted, preventing concentration and preserving non-degenerate gradients. We prove two main results: (1) a trainability lower bound for LieTrunc-QNN, and (2) that the Fubini-Study metric rank is bounded by the algebraic span of generators, showing expressivity is governed by structure rather than parameter count. Compact Lie subalgebras also provide inherent robustness to perturbations. Importantly, we establish a polynomial trainability regime where gradient variance decays polynomially instead of exponentially. Experiments (n=2-6) validate the theory: LieTrunc-QNN maintains stable gradients and high effective dimension, while random truncation leads to metric rank collapse. At n=6, full metric rank is preserved (rank=16). Results support a scaling law between gradient variance and effective dimension. This work provides a unified geometric framework for QNN design, linking Lie algebra, manifold geometry, and optimization.

arXiv:2604.02697•Apr 3, 2026•0 citations

Circuit Synthesis for Variational Quantum Algorithms: The Depth of Algebra and Geometry

Boris Tsvelikhovskiy

The analysis of Hamiltonian Variational Ansätze centers on three fundamental questions: which states are exactly reachable from a given initial state, what native control depth suffices for local and global synthesis, and whether a bounded circuit extension can expose a direction of decreasing cost at a suboptimal output. We address these challenges by leveraging the dynamical Lie algebras (DLAs) associated with the circuits and the geometry of the initial state's orbit under the action of the corresponding dynamical Lie group $G$. We show that an initial state has a closed projective orbit whenever it is a nondegenerate eigenstate of a control Hamiltonian. Combined with asymptotic QAOA convergence, this guarantees exact, finite depth ground state preparation for architectures employing diagonal problem Hamiltonians and irreducible stoquastic mixers. We bound the number of native control exponentials required for local unitary synthesis and state preparation using DLA commutator filtration growth alongside reachable orbit dimensions. Complementing these local results, we construct global frameworks on invariant sectors equipped with full unitary or special unitary dynamical Lie algebras. These constructions yield explicit applications to Pauli-generated systems and constrained maximum independent set multi-angle QAOA circuits; specifically, we show that their native control costs are directly determined by commutator filtration depth and independent-set toggle-group geometry, respectively. Finally, if the dynamical Lie group is compact, the observable $H_P$ generates an admissible evolution ($iH_P \in \mathfrak{g}$, where $\mathfrak{g}$ is the DLA), and the projective orbit of the initial state contains a ground state ray of $H_P$, then a bounded depth, identity-initialized circuit extension exhibits either first-order descent or negative curvature at every suboptimal circuit output.

quant-pharXiv:2610.22689•Oct 19, 2026•0 citations
Orbit dimension, not parameter count, bounds quantum sample complexity. For continuous parameterized quantum circuits on nnn qubits, the minimax estimation risk obeys [ \mathcal{R}(d_{\mathrm{eff}}, M) ;\geq; \Omega!\left(\sqrt{\frac{d_{\mathrm{eff}}}{M}}\right), ] where deff=dim⁡(OO)d_{\mathrm{eff}} = \dim(\mathcal{O}_O)deff​=dim(OO​) is the real dimension of the observable's adjoint orbit under the dynamical Lie group. For balanced Pauli observables O=Z⊗nO = Z^{\otimes n}O=Z⊗n, quotienting the dynamical Lie algebra by the stabilizer subgroup yields the complex Grassmannian Gr(2n−1,2n)Gr(2^{n-1}, 2^n)Gr(2n−1,2n) of real dimension deff=22n−1=124n≈12dim⁡(g)d_{\mathrm{eff}} = 2^{2n-1} = \frac{1}{2}4^n \approx \frac{1}{2}\dim(\mathfrak{g})deff​=22n−1=21​4n≈21​dim(g), which matches upper bounds of Caro et al.(2022) and Ohno(2025) up to a universal constant and resolves the exact tightness rate Θ(dim⁡(g)/M)\Theta(\sqrt{\dim(\mathfrak{g})/M})Θ(dim(g)/M​). For rank-1 projective observables, the orbit contracts instead to complex projective space CP2n−1\mathbb{CP}^{2^n-1}CP2n−1 with dimension deff=2(2n−1)d_{\mathrm{eff}} = 2(2^n-1)deff​=2(2n−1). Sample requirements therefore explode exponentially with qubit count. The scaling identifies Wang & Wu's (2025) dimension-independent O(1/M)\mathcal{O}(1/\sqrt{M})O(1/M​) uniform convergence bound as an artifact of uniform stability rather than true estimation power. PennyLane benchmarks across n=1,2,3,4n=1,2,3,4n=1,2,3,4 qubits confirm this exponential threshold. Barren plateaus and statistical sample bottlenecks stem from an identical geometric origin: volume growth in SU(2n)SU(2^n)SU(2n) simultaneously suppresses gradient variances as ∼1/dim⁡(g)\sim 1/\dim(\mathfrak{g})∼1/dim(g) and drives sample complexity up to ∼dim⁡(g)/M\sim \sqrt{\dim(\mathfrak{g})/M}∼dim(g)/M​.
Quantum learnability is an algebraic invariant. While optimization and statistical generalization have long been analyzed through disconnected mathematical formalisms in quantum machine learning, both phenomena emerge as dual geometric projections of the dynamical Lie algebra (DLA) dimension g = Lie{-i H_k} ⊆ su(2^n); across any parameterized circuit family, polynomial trainability (Var[∇C] ≥ Ω(1/poly(n))) and polynomial generalizability (M*(ε) ≤ O(poly(n)/ε^2)) are strictly equivalent and hold if and only if dim(g_n) ∈ O(poly(n)). For any observable with adjoint orbit O_O, the operational product satisfies Var[∇C] · M*(ε) = Θ(dim(O_O) / (dim(g_n) ε^2)), which collapses to the universal, dimension-free invariant Θ(1/ε^2) for all generic, non-degenerate observables, identically independent of system size n and algebraic dimension dim(g_n). Unconstrained expressivity (dim(g_n) ~ 4^n) triggers a simultaneous catastrophe: exponential gradient vanishing accompanied by exponential sample complexity explosion. Subalgebra confinement cures both failures. By restricting generators to permutation-equivariant or symmetry-protected subalgebras, the circuit enforces dim(g_n) ≤ O(poly(n)) and secures non-vanishing gradient flow alongside statistical sample efficiency. Exact statevector simulations in PennyLane across n ∈ {2, 3, 4, 5, 6} corroborate this duality: hardware-efficient ansätze exhibit a 17.22× variance collapse, whereas symmetry-protected models sustain Θ(1) gradients and achieve twofold lower test error.
Understanding how the performance of parameterized quantum circuits scales with available resources is important for characterizing their trainability and effective model capacity. In this study, we numerically investigate data scaling, model scaling, and compute scaling in parameterized quantum circuits and examine Lie-algebraic quantities as alternative measures of model size. In addition to the number of circuit parameters, we consider the dimension of the dynamical Lie algebra, the observable-orbit dimension, and a Jacobian effective dimension defined as the rank of the Jacobian of the parameterized observable orbit. Using a regression task with randomly generated Pauli-string generators, we observe decreasing loss with increasing training dataset size, parameter size, and number of optimization iterations over the ranges investigated. For model scaling, the dynamical Lie algebra and observable orbit dimensions rapidly saturate as the parameter size increases, whereas the Jacobian effective dimension remains strongly correlated with the parameter size and exhibits comparable scaling behavior. These results suggest that the Jacobian effective dimension provides a geometry-aware measure of the locally accessible observable degrees of freedom of finite-depth parameterized quantum circuits and may serve as a useful scaling parameter beyond the nominal parameter size.
Commutation failure controls quantum generalization. When data encoding gates fail to commute with trainable parameter gates across consecutive re-uploading layers, the effective dynamical Lie algebra becomes data-dependent; this dependency invalidates the parameter-isolation assumption underpinning existing Lie-algebraic frameworks (Caro et al., 2022; Ohno, 2025). We resolve this structural gap by proving that the statistical generalization capacity of L-layer data re-uploading circuits is tightly bounded by the effective hypothesis dimension: d_eff(L) = min(L * p_layer, |Ω_L|) <= min(L * p_layer, (2L + 1)^d), where d is the input data dimension, p_layer the number of trainable parameters per layer, and |Ω_L| the cardinality of accessible Fourier frequencies generated through non-commuting Lie brackets in the joint dynamical Lie algebra g_joint = Lie_R{i H_j, i G_k}. Harmonic analysis combined with metric entropy chaining establishes matching upper and lower bounds; together, they settle the exact minimax estimation rate R*(F_L, M) = \tilde{\Theta}(\sqrt{d_eff(L)/M}). Two distinct orbit concepts govern learning. The point-wise adjoint orbit dimension dim(O_O) <= dim(g_joint) at fixed x quantifies single-point state variation, whereas the global function-space Fourier capacity |Ω_L| governs sample complexity across the full data domain; conflating these quantities yields vacuous bounds. An initial basis rotation unlocks harmonic modes up to k_max = L; without this rotation, eigenstate initialization constrains the cutoff to k_max = L - 1. Bandwidth dictates depth. For target functions with harmonic bandlimit k_target, the network underfits when L < k_target and achieves full expressivity convergence once L >= k_target. Permutation-equivariant data encoding compresses the joint DLA to an su(2) subalgebra of dimension 3. This algebraic confinement suppresses the 4^n - 1 dimensional explosion across arbitrary qubit counts and shields the circuit from the curse of dimensionality. Numerical benchmarks in PennyLane across L in {1, 2, 4, 8} and sample sizes M in [10, 200] corroborate this harmonic-algebraic scaling and its associated uniform convergence rates.
Scalable parameterized quantum circuits for machine learning face three obstacles: barren plateaus, no guarantee the learned function class is classically hard, and prohibitive circuit evaluations per gradient step. We propose the unitary brick-wall: a kkk-particle fermionic architecture for nearest-neighbor hardware, combining Reconfigurable Beam Splitters with interleaved single-qubit phases and a non-Gaussian magic-state encoding, where kkk trades classical hardness against training cost. Trainable. The brick-wall has dynamical Lie algebra u(n)\mathfrak{u}(n)u(n) and is surjective onto U(n)U(n)U(n) via Givens rotations, enabling Haar initialization. Two-body correlator readouts achieve gradient variance Θ(k3/n5)Θ(k^3/n^5)Θ(k3/n5), polynomial in nnn throughout n−2k=Ω(n)n-2k=Ω(n)n−2k=Ω(n). Expressive. Hardness is controlled by kkk: best-known classical sampling runs in time 2Θ(k)2^{Θ(k)}2Θ(k), worst-case #P-hardness holds from k=nεk=n^εk=nε, and the average-case machinery of Fermion Sampling applies at k=Θ(n)k=Θ(n)k=Θ(n). The sampler is input-dependent, so costs are paid per query: at our operating point k=100k=100k=100 one query needs ∼1015\sim 10^{15}∼1015 operations, beyond commercial query rates, and at k=120k=120k=120 an exascale machine sustains under one query per second. Efficient. A multi-layer parallel parameter-shift rule computes all O(n2)O(n^2)O(n2) gradients from 4kn4kn4kn circuit evaluations per step, a factor n/kn/kn/k reduction over the standard rule's 4n24n^24n2, growing linearly with nnn at fixed kkk. The unitary butterfly targets all-to-all hardware, with 2log⁡n2\log n2logn layers of parallel gates and (3/2)nlog⁡n(3/2)n\log n(3/2)nlogn parameters, similar hardness guarantees, and 4klog⁡n4k\log n4klogn evaluations per step -- the same factor-n/kn/kn/k reduction. Its trainability holds at two levels: absence of exponential barren plateaus is unconditional, while the sharp Θ(k3/n5)Θ(k^3/n^5)Θ(k3/n5) rate needs a conjectured scalar comparison with the two-particle Haar second moment.
Quantum Machine Learning (QML) is fundamentally limited by two challenges: barren plateaus (exponentially vanishing gradients) and the fragility of parameterized quantum circuits under noise. Despite extensive empirical studies, a unified theoretical framework remains lacking. We introduce LieTrunc-QNN, an algebraic-geometric framework that characterizes trainability via Lie-generated dynamics. Parameterized quantum circuits are modeled as Lie subalgebras of u(2^n), whose action induces a Riemannian manifold of reachable quantum states. Expressivity is reinterpreted as intrinsic manifold dimension and geometry. We establish a geometric capacity-plateau principle: increasing effective dimension leads to exponential gradient suppression due to concentration of measure. By restricting to structured Lie subalgebras (LieTrunc), the manifold is contracted, preventing concentration and preserving non-degenerate gradients. We prove two main results: (1) a trainability lower bound for LieTrunc-QNN, and (2) that the Fubini-Study metric rank is bounded by the algebraic span of generators, showing expressivity is governed by structure rather than parameter count. Compact Lie subalgebras also provide inherent robustness to perturbations. Importantly, we establish a polynomial trainability regime where gradient variance decays polynomially instead of exponentially. Experiments (n=2-6) validate the theory: LieTrunc-QNN maintains stable gradients and high effective dimension, while random truncation leads to metric rank collapse. At n=6, full metric rank is preserved (rank=16). Results support a scaling law between gradient variance and effective dimension. This work provides a unified geometric framework for QNN design, linking Lie algebra, manifold geometry, and optimization.
The analysis of Hamiltonian Variational Ansätze centers on three fundamental questions: which states are exactly reachable from a given initial state, what native control depth suffices for local and global synthesis, and whether a bounded circuit extension can expose a direction of decreasing cost at a suboptimal output. We address these challenges by leveraging the dynamical Lie algebras (DLAs) associated with the circuits and the geometry of the initial state's orbit under the action of the corresponding dynamical Lie group GGG. We show that an initial state has a closed projective orbit whenever it is a nondegenerate eigenstate of a control Hamiltonian. Combined with asymptotic QAOA convergence, this guarantees exact, finite depth ground state preparation for architectures employing diagonal problem Hamiltonians and irreducible stoquastic mixers. We bound the number of native control exponentials required for local unitary synthesis and state preparation using DLA commutator filtration growth alongside reachable orbit dimensions. Complementing these local results, we construct global frameworks on invariant sectors equipped with full unitary or special unitary dynamical Lie algebras. These constructions yield explicit applications to Pauli-generated systems and constrained maximum independent set multi-angle QAOA circuits; specifically, we show that their native control costs are directly determined by commutator filtration depth and independent-set toggle-group geometry, respectively. Finally, if the dynamical Lie group is compact, the observable HPH_PHP​ generates an admissible evolution (iHP∈giH_P \in \mathfrak{g}iHP​∈g, where g\mathfrak{g}g is the DLA), and the projective orbit of the initial state contains a ground state ray of HPH_PHP​, then a bounded depth, identity-initialized circuit extension exhibits either first-order descent or negative curvature at every suboptimal circuit output.