arXiv subjects

Srinivasan Arunachalam

Publications and source records attributed to Srinivasan Arunachalam.

At least 19 recordsLinked to original sources

arXiv2026

Marton's conjecture in polynomial time

Gowers, Green, Manners, and Tao (Annals '25) recently resolved Marton's polynomial Freiman-Ruzsa conjecture. We give an algorithmic counterpart to their result: given uniform sampling and membership-oracle access to a set $A \subseteq \mathbb{F}_2^n$ with doubling constant at most $K$, our algorithm outputs a subspace of size at most $|A|$ whose $K^{O(1)}$ translates cover $A$. The algorithm runs in $\textsf{poly}(n,K)$ time. As applications, we obtain polynomial-time algorithms for a variety of learning problems, including quadratic Goldreich-Levin, improper agnostic tomography of stabilizer states, and tomography of quantum states with bounded stabilizer extent.

cs.CC↗

arXiv2026

A note on the partition bound for one-way classical communication complexity

We present a linear program for a one-way version of the partition bound of Jain and Klauck, denoted $\mathsf{prt}^1_\varepsilon(f)$, and show that it characterizes the one-way randomized communication complexity $\mathsf{R}_\varepsilon^1(f)$ with shared randomness and worst-case error $\varepsilon$ of any relation $f\subseteq\mathcal{X}\times\mathcal{Y}\times\mathcal{Z}$. More precisely, for all $\varepsilon,δ\in(0,1/2)$, $\mathsf{R}_\varepsilon^1(f) \geq \log\mathsf{prt}_\varepsilon^1(f)$ and $\mathsf{R}_{\varepsilon+δ}^1(f) \leq \lceil\log\mathsf{prt}_\varepsilon^1(f) + \log\log(1/δ)\rceil$. This improves upon the characterization of $\mathsf{R}_\varepsilon^1(f)$ in terms of the rectangle bound due to Jain, Klauck, and Nayak [STOC'08], which is tight only up to an additive $O(\log(1/δ))$ term.

cs.CC↗

arXiv2026

Separating quantum circuits from classical LLMs

Modern large language models - transformers and diffusion language models - are built around two canonical algorithmic tasks: prediction and generation. We prove unconditional separations between low-depth quantum computation and the corresponding bounded-resource classical language-model architectures in both regimes. Concretely, we exhibit the following: 1. Distributional separation. We give a distribution that is sampleable by $\textsf{QNC}^0$ circuits (i.e., a family of constant-depth quantum circuits consisting of bounded fan-in gates) that no constant-round diffusion language model ($\textsf{DLM}$) with shallow scheduling and denoising can sample within constant distance, even when allowed sublinear chain-of-thought and output-token revision/remasking events, the very features modern $\textsf{DLM}$s rely on. 2. Functional separation. We exhibit a function computable in $\land \circ \textsf{QNC}^0[\log\log n]$ (i.e., a family of O$(\log\log n)$-depth $\textsf{QNC}^0$ circuits, where $n$ is the input length, followed by a single classical $\mathsf{AND}$ gate) such that any constant-depth decoder-only transformer computing the function must be large: it would have to have width $n^{Ω(1)}$. Together, our work initiates the study of quantum advantage in the era of large language models.

quant-ph↗

arXiv2026

No low-degree tests for quantum states

We study the problem of testing low-degree phase states, namely m-qudit quantum states of the form $q^{-m/2} \sum_{x \in \mathbb{F}_q^m} ω^{f(x)} |x>$, where $f$ is a degree-$d$ polynomial. In contrast to the classical setting, where low-degree polynomials admit highly efficient classical testers, it is not known whether analogous quantum tests exist. We show that no such quantum low-degree test exists: any tester requires $Ω(\binom{\lfloor m/2\rfloor}{\lfloor (d-1)/2 \rfloor})$ copies to determine whether a given state is a degree-$d$ phase state or is far from every such state. Our results follow from a general framework that relates quantum testing of codeword phase states to classical decoding properties of the dual code, which allows us to leverage known bounds on the tolerance of high-rate Reed--Muller codes to random errors.

quant-ph↗

arXiv2026

Optimal Stabilizer Testing and Learning with Limited Quantum Memory

We study stabilizer state testing and learning with limited coherent quantum memory. Here an algorithm sequentially receives copies of an unknown $n$-qubit state, but may keep only $k$ qubits of coherent quantum memory between measurements. With unrestricted memory, seminal work of Gross, Nezami and Walter showed how to test $n$-qubit stabilizer states using $6$ copies, which is dimension independent, unlike the learning complexity of $Θ(n)$. We show that this testing-vs-learning separation is lost under memory constraints. More concretely we show that (1) The sample complexity of testing stabilizer states in the $k$-qubit memory framework is $Θ(n-k)$. Our upper bound goes via a novel connection to the hidden shift problem and the lower bound is proven using a novel approach to average case bounds on likelihood ratios via combinatorics of the stochastic orthogonal group. (2) The sample complexity of learning stabilizer states with $k$ qubits of memory, in the non-adaptive framework, is $Θ(n^2/k)$. As a further application of our techniques, we prove an exponential lower bound for purity testing even when the memory may be left coherent throughout the protocol. Our main results identify coherent quantum memory as the resource enabling the usual separation between stabilizer testing and learning. In particular, even with $k=0.99n$ qubits of memory, there is no constant-copy stabilizer tester; furthermore for $k=cn$ qubits of memory (for $0< c < 1$), stabilizer testing is as hard as learning, with both requiring $Θ(n)$ copies.

quant-ph↗

arXiv2026

Tomography of quantum states with bounded extent

We give a general framework for tomography of states that have bounded-extent with respect to a structured class of states. Let $\textsf{C}$ be a family of $n$-qubit states such that: $(i)$ $\textsf{C}$ is succinctly representable and $(ii)$ there is a weak agnostic learner of $\textsf{C}$. We give a tomography protocol for an unknown state $|ψ\rangle$ that is promised to admit a decomposition of the form $|ψ\rangle = \sum_i c_i |ϕ_i\rangle$, where $|ϕ_i\rangle \in \textsf{C}$ with bounded $\ell_1$-norm of the coefficients (which we call extent). Our main contribution is to show that a weak agnostic learner for $\textsf{C}$ can be boosted into a tomography algorithm for states with bounded extent with respect to $\textsf{C}$. Our reduction is black-box and applies broadly across model classes. As an application, when $\textsf{C}$ is the class of stabilizer states, we obtain tomography algorithms for states with stabilizer extent $ξ$ up to trace distance $\varepsilon$, in time $\textsf{poly}(n,(ξ/\varepsilon)^{\log(ξ/\varepsilon)})$, which is improvable to $ \textsf{poly}(n,ξ,1/\varepsilon)$ assuming the algorithmic polynomial Freiman-Ruzsa conjecture in the high-doubling regime. When the unknown state $|ψ\rangle$ is arbitrary, we give an algorithmic decomposition result in the spirit of a weak regularity lemma for quantum states with respect to $\textsf{C}$ and show that the structure in $|ψ\rangle$ that is explainable by $\textsf{C}$ can be efficiently learned. Our main conceptual message is that agnostic learning of a structured base class automatically yields learnability of its low-complexity linear span.

quant-ph↗

arXiv2026

An algorithmic Polynomial Freiman-Ruzsa theorem

We provide algorithmic versions of the Polynomial Freiman-Ruzsa theorem of Gowers, Green, Manners, and Tao (Ann. of Math., 2025). In particular, we give a polynomial-time algorithm that, given a set $A \subseteq \mathbb{F}_2^n$ with doubling constant $K$, returns a subspace $V \subseteq \mathbb{F}_2^n$ of size $|V| \leq |A|$ such that $A$ can be covered by $2K^C$ translates of $V$, for a universal constant $C>1$. We also provide efficient algorithms for several "equivalent" formulations of the Polynomial Freiman-Ruzsa theorem, such as the polynomial Gowers inverse theorem, the classification of approximate Freiman homomorphisms, and quadratic structure-vs-randomness decompositions. Our algorithmic framework is based on a new and optimal version of the Quadratic Goldreich-Levin algorithm, which we obtain using ideas from quantum learning theory. This framework fundamentally relies on a connection between quadratic Fourier analysis and symplectic geometry, first speculated by Green and Tao (Proc. of Edinb. Math. Soc., 2008) and which we make explicit in this paper.

math.CO↗

arXiv2026

Learning depth-3 circuits via quantum agnostic boosting

We initiate the study of quantum agnostic learning of phase states with respect to a function class $\mathsf{C}\subseteq \{c:\{0,1\}^n\rightarrow \{0,1\}\}$: given copies of an unknown $n$-qubit state $|ψ\rangle$ which has fidelity $\textsf{opt}$ with a phase state $|ϕ_c\rangle=\frac{1}{\sqrt{2^n}}\sum_{x\in \{0,1\}^n}(-1)^{c(x)}|x\rangle$ for some $c\in \mathsf{C}$, output $|ϕ\rangle$ which has fidelity $|\langle ϕ| ψ\rangle|^2 \geq \textsf{opt}-\varepsilon$. To this end, we give agnostic learning protocols for the following classes: (i) Size-$t$ decision trees which runs in time $\textsf{poly}(n,t,1/\varepsilon)$. This also implies $k$-juntas can be agnostically learned in time $\textsf{poly}(n,2^k,1/\varepsilon)$. (ii) $s$-term DNF formulas in time $\textsf{poly}(n,(s/\varepsilon)^{\log \log (s/\varepsilon) \cdot \log(1/\varepsilon)})$. Our main technical contribution is a quantum agnostic boosting protocol which converts a weak agnostic learner, which outputs a parity state $|ϕ\rangle$ such that $|\langle ϕ|ψ\rangle|^2\geq \textsf{opt}/\textsf{poly}(n)$, into a strong learner which outputs a superposition of parity states $|ϕ'\rangle$ such that $|\langle ϕ'|ψ\rangle|^2\geq \textsf{opt} - \varepsilon$. Using quantum agnostic boosting, we obtain a $n^{O(\log(n/\varepsilon) \cdot \log \log n)}$-time algorithm for $\varepsilon$-learning $\textsf{poly}(n)$-sized depth-$3$ circuits (consisting of $\textsf{AND}$, $\textsf{OR}$, $\textsf{NOT}$ gates) in the uniform $\textsf{PAC}$ model given quantum examples. Classically, obtaining an algorithm with a similar complexity has been an open question in the $\textsf{PAC}$ model and our work answers this given quantum examples.

quant-ph↗

arXiv2025

Trade-offs between Entanglement and Communication

We study the advantages of quantum communication models over classical communication models that are equipped with a limited number of qubits of entanglement. In this direction, we give explicit partial functions on $n$ bits for which reducing the entanglement increases the classical communication complexity exponentially. Our separations are as follows. For every $k\ge 1$: $Q\|^*$ versus $R2^*$: We show that quantum simultaneous protocols with $\tildeΘ(k^5 \log^3 n)$ qubits of entanglement can exponentially outperform two-way randomized protocols with $O(k)$ qubits of entanglement. This resolves an open problem from [Gav08] and improves the state-of-the-art separations between quantum simultaneous protocols with entanglement and two-way randomized protocols without entanglement [Gav19, GRT22]. $R\|^*$ versus $Q\|^*$: We show that classical simultaneous protocols with $\tildeΘ(k \log n)$ qubits of entanglement can exponentially outperform quantum simultaneous protocols with $O(k)$ qubits of entanglement, resolving an open question from [GKRW06, Gav19]. The best result prior to our work was a relational separation against protocols without entanglement [GKRW06]. $R\|^*$ versus $R1^*$: We show that classical simultaneous protocols with $\tildeΘ(k\log n)$ qubits of entanglement can exponentially outperform randomized one-way protocols with $O(k)$ qubits of entanglement. Prior to our work, only a relational separation was known [Gav08]. Our techniques can also be used to show advantages of quantum communication models over hybrid classical-quantum models, i.e., models that have a large amount of both classical communication and quantum simultaneous communication.

quant-ph↗

arXiv2025

Learning stabilizer structure of quantum states

We consider the task of learning a structured stabilizer decomposition of an arbitrary $n$-qubit quantum state $|ψ\rangle$: for $ε> 0$, output a state $|ϕ\rangle$ with stabilizer-rank $\textsf{poly}(1/ε)$ such that $|ψ\rangle=|ϕ\rangle+|ϕ'\rangle$ where $|ϕ'\rangle$ has stabilizer fidelity $< ε$. We first show the existence of such decompositions using the recently established inverse theorem for the Gowers-$3$ norm of states [AD,STOC'25]. To learn this structure, we initiate the task of self-correction of a state $|ψ\rangle$ with respect to a class of states $S$: given copies of $|ψ\rangle$ which has fidelity $\geq τ$ with a state in $S$, output $|ϕ\rangle \in S$ with fidelity $|\langle ϕ| ψ\rangle|^2 \geq τ^C$ for a constant $C>1$. Assuming the algorithmic polynomial Frieman-Rusza (APFR) conjecture in the high doubling regime (whose combinatorial version was recently resolved [GGMT,Annals of Math.'25]), we give a polynomial-time algorithm for self-correction of stabilizer states. Given access to the state preparation unitary $U_ψ$ for $|ψ\rangle$ and its controlled version $cU_ψ$, we give a polynomial-time protocol that learns a structured decomposition of $|ψ\rangle$. Without assuming APFR, we give a quasipolynomial-time protocol for the same task. As our main application, we give learning algorithms for states $|ψ\rangle$ promised to have stabilizer extent $ξ$, given access to $U_ψ$ and $cU_ψ$. We give a protocol that outputs $|ϕ\rangle$ which is constant-close to $|ψ\rangle$ in time $\textsf{poly}(n,ξ^{\log ξ})$, which can be improved to polynomial-time assuming APFR. This gives an unconditional learning algorithm for stabilizer-rank $k$ states in time $\textsf{poly}(n,k^{k^2})$. As far as we know, learning arbitrary states with even stabilizer-rank $2$ was unknown.

quant-ph↗

arXiv2025

Algorithmic Polynomial Freiman-Ruzsa Theorems

We prove algorithmic versions of the polynomial Freiman-Ruzsa theorem of Gowers, Green, Manners, and Tao (Annals of Mathematics, 2025) in additive combinatorics. In particular, we give classical and quantum polynomial-time algorithms that, for $A \subseteq \mathbb{F}_2^n$ with doubling constant $K$, learn an explicit description of a subspace $V \subseteq \mathbb{F}_2^n$ of size $|V| \leq |A|$ such that $A$ can be covered by $K^C$ translates of $V$, for a universal constant $C>1$.

math.CO↗

arXiv2025

Testing and learning structured quantum Hamiltonians

We consider the problems of testing and learning an unknown $n$-qubit Hamiltonian $H$ from queries to its evolution operator $e^{-iHt}$ under the normalized Frobenius norm. We prove: 1. Local Hamiltonians: We give a tolerant testing protocol to decide if $H$ is $ε_1$-close to $k$-local or $ε_2$-far from $k$-local, with $O(1/(ε_2-ε_1)^{4})$ queries, solving open questions posed in a recent work by Bluhm et al. For learning a $k$-local $H$ up to error $ε$, we give a protocol with query complexity $\exp(O(k^2+k\log(1/ε)))$ independent of $n$, by leveraging the non-commutative Bohnenblust-Hille inequality. 2. Sparse Hamiltonians: We give a protocol to test if $H$ is $ε_1$-close to being $s$-sparse (in the Pauli basis) or $ε_2$-far from being $s$-sparse, with $O(s^{6}/(ε_2^2-ε_1^2)^{6})$ queries. For learning up to error $ε$, we show that $O(s^{4}/ε^{8})$ queries suffice. 3. Learning without memory: The learning results stated above have no dependence on $n$, but require $n$-qubit quantum memory. We give subroutines that allow us to learn without memory; increasing the query complexity by a $(\log n)$-factor in the local case and an $n$-factor in the sparse case. 4. Testing without memory: We give a new subroutine called Pauli hashing, which allows one to tolerantly test $s$-sparse Hamiltonians with $O(s^{14}/(ε_2^2-ε_1^2)^{18})$ queries. A key ingredient is showing that $s$-sparse Pauli channels can be tolerantly tested under the diamond norm with $O(s^2/(ε_2-ε_1)^6)$ queries. Along the way, we prove new structural theorems for local and sparse Hamiltonians. We complement our learning results with polynomially weaker lower bounds. Furthermore, our algorithms use short time evolutions and do not assume prior knowledge of the terms in the support of the Pauli spectrum.

quant-ph↗

arXiv2024

Polynomial-time tolerant testing stabilizer states

We consider the following task: suppose an algorithm is given copies of an unknown $n$-qubit quantum state $|ψ\rangle$ promised $(i)$ $|ψ\rangle$ is $\varepsilon_1$-close to a stabilizer state in fidelity or $(ii)$ $|ψ\rangle$ is $\varepsilon_2$-far from all stabilizer states, decide which is the case. We show that for every $\varepsilon_1>0$ and $\varepsilon_2\leq \varepsilon_1^C$, there is a $\textsf{poly}(1/\varepsilon_1)$-sample and $n\cdot \textsf{poly}(1/\varepsilon_1)$-time algorithm that decides which is the case (where $C>1$ is a universal constant). Our proof includes a new definition of Gowers norm for quantum states, an inverse theorem for the Gowers-$3$ norm of quantum states and new bounds on stabilizer covering for structured subsets of Paulis using results in additive combinatorics.

quant-ph↗

arXiv2024

Matrix hypercontractivity, streaming algorithms and LDCs: the large alphabet case

We prove a hypercontractive inequality for matrix-valued functions defined over large alphabets. In order to do so, we prove a generalization of the powerful $2$-uniform convexity inequality for trace norms of Ball, Carlen, Lieb (Inventiones Mathematicae'94). Using our hypercontractive~inequality, we present upper and lower bounds for the communication complexity of the Hidden Hypermatching problem defined over large alphabets. We then consider streaming algorithms for approximating the value of Unique Games on a hypergraph with $t$-size hyperedges. By using our communication lower bound, we show that every streaming algorithm in the adversarial model achieving an $(r-\varepsilon)$-approximation of this value requires $Ω(n^{1-2/t})$ quantum space, where $r$ is the alphabet size. We next present a lower bound for locally decodable codes (LDC) $\mathbb{Z}_r^n\to \mathbb{Z}_r^N$ over large alphabets with recoverability probability at least $1/r + \varepsilon$. Using hypercontractivity, we give an exponential lower bound $N = 2^{Ω(\varepsilon^4 n/r^4)}$ for $2$-query (possibly non-linear) LDCs over $\mathbb{Z}_r$ and using the non-commutative Khintchine inequality we prove an improved lower bound of $N = 2^{Ω(\varepsilon^2 n/r^2)}$.

cs.CC↗

arXiv2024

A note on polynomial-time tolerant testing stabilizer states

We show an improved inverse theorem for the Gowers-$3$ norm of $n$-qubit quantum states $|ψ\rangle$ which states that: for every $γ\geq 0$, if the $\textsf{Gowers}(|ψ\rangle,3)^8 \geq γ$ then the stabilizer fidelity of $|ψ\rangle$ is at least $γ^C$ for some constant $C>1$. This implies a constant-sample polynomial-time tolerant testing algorithm for stabilizer states which accepts if an unknown state is $\varepsilon_1$-close to a stabilizer state in fidelity and rejects when $|ψ\rangle$ is $\varepsilon_2 \leq \varepsilon_1^C$-far from all stabilizer states, promised one of them is the case.

quant-ph↗

arXiv2024

Distributed inner product estimation with limited quantum communication

We consider the task of distributed inner product estimation when allowed limited quantum communication. Here, Alice and Bob are given $k$ copies of an unknown $n$-qubit quantum states $\vert ψ\rangle,\vert ϕ\rangle$ respectively. They are allowed to communicate $q$ qubits and unlimited classical communication, and their goal is to estimate $|\langle ψ|ϕ\rangle|^2$ up to constant accuracy. We show that $k=Θ(\sqrt{2^{n-q}})$ copies are essentially necessary and sufficient for this task (extending the work of Anshu, Landau and Liu (STOC'22) who considered the case when $q=0$). Additionally, we consider estimating $|\langle ψ|M|ϕ\rangle|^2$, for arbitrary Hermitian $M$. For this task we show that certain norms on $M$ characterize the sample complexity of estimating $|\langle ψ|M|ϕ\rangle|^2$ when using only classical~communication.

quant-ph↗

arXiv2024

Learning low-degree quantum objects

We consider the problem of learning low-degree quantum objects up to $\varepsilon$-error in $\ell_2$-distance. We show the following results: $(i)$ unknown $n$-qubit degree-$d$ (in the Pauli basis) quantum channels and unitaries can be learned using $O(1/\varepsilon^d)$ queries (independent of $n$), $(ii)$ polynomials $p:\{-1,1\}^n\rightarrow [-1,1]$ arising from $d$-query quantum algorithms can be classically learned from $O((1/\varepsilon)^d\cdot \log n)$ many random examples $(x,p(x))$ (which implies learnability even for $d=O(\log n)$), and $(iii)$ degree-$d$ polynomials $p:\{-1,1\}^n\to [-1,1]$ can be learned through $O(1/\varepsilon^d)$ queries to a quantum unitary $U_p$ that block-encodes $p$. Our main technical contributions are new Bohnenblust-Hille inequalities for quantum channels and completely bounded~polynomials.

quant-ph↗

arXiv2023

On the Role of Entanglement and Statistics in Learning

In this work we make progress in understanding the relationship between learning models with access to entangled, separable and statistical measurements in the quantum statistical query (QSQ) model. To this end, we show the following results. $\textbf{Entangled versus separable measurements.}$ The goal here is to learn an unknown $f$ from the concept class $C\subseteq \{f:\{0,1\}^n\rightarrow [k]\}$ given copies of $\frac{1}{\sqrt{2^n}}\sum_x \vert x,f(x)\rangle$. We show that, if $T$ copies suffice to learn $f$ using entangled measurements, then $O(nT^2)$ copies suffice to learn $f$ using just separable measurements. $\textbf{Entangled versus statistical measurements}$ The goal here is to learn a function $f \in C$ given access to separable measurements and statistical measurements. We exhibit a class $C$ that gives an exponential separation between QSQ learning and quantum learning with entangled measurements (even in the presence of noise). This proves the "quantum analogue" of the seminal result of Blum et al. [BKW'03]. that separates classical SQ and PAC learning with classification noise. $\textbf{QSQ lower bounds for learning states.}$ We introduce a quantum statistical query dimension (QSD), which we use to give lower bounds on the QSQ learning. With this we prove superpolynomial QSQ lower bounds for testing purity, shadow tomography, Abelian hidden subgroup problem, degree-$2$ functions, planted bi-clique states and output states of Clifford circuits of depth $\textsf{polylog}(n)$. $\textbf{Further applications.}$ We give and $\textit{unconditional}$ separation between weak and strong error mitigation and prove lower bounds for learning distributions in the QSQ model. Prior works by Quek et al. [QFK+'22], Hinsche et al. [HIN+'22], and Nietner et al. [NIS+'23] proved the analogous results $\textit{assuming}$ diagonal measurements and our work removes this assumption.

quant-ph↗