Low-Degree Indistinguishability

Updated 16 January 2026

Low-degree indistinguishability is a framework that assesses the inability of bounded-complexity tests to distinguish complex objects like quantum states, graphs, and distributions.
It employs metrics such as low-degree likelihood ratios, fringe visibility, and homomorphism counts to rigorously quantify indistinguishability across quantum, combinatorial, and complexity domains.
Practical implications include enhanced quantum metrology, cryptographic hardness proofs, and refined structural insights in graph theory, backed by experimental validations.

Low-degree indistinguishability is a multi-faceted concept central to quantum information, average-case complexity, quantum optics, and combinatorics. Broadly, it characterizes the extent to which two objects—particle states, probability distributions, graphs, or percolation clusters—remain indistinguishable when probed by statistical or algebraic tests of bounded complexity or polynomial degree. This framework enables rigorous analysis of quantum coherence resources, threshold phenomena in planted vs null models, the fundamental limits of distinguishing symmetric distributions, and the structure of indistinguishability classes in probabilistic combinatorics and graph theory. Recent advances have formalized, measured, and exploited low-degree indistinguishability in experimental and theoretical contexts, underpinning practical quantum metrology, cryptographic hardness results, and structural graph properties.

1. Theoretical Foundations: Indistinguishability by Low-Degree Tests

Low-degree indistinguishability often refers to the inability of bounded-complexity statistical or algebraic tests to differentiate between two objects. In quantum information (e.g., photon states), classical probability, and random matrix theory, "low-degree" typically references observables expressible as polynomials of the system parameters up to degree $D$ , or marginal tests up to $k$ variables (Hsieh et al., 9 Jan 2026, Williamson, 2021).

Formally, given two distributions $P$ (planted) and $N$ (null) on a space $\Omega$ , their degree- $D$ indistinguishability is quantified via the low-degree likelihood ratio (LDLR):

$L^{\leq D}(x) = \text{Projection of } \frac{dP}{dN}(x) \text{ onto degree-}\leq D \text{ polynomials}$

$\chi^2_{\leq D}(P\|N) := \mathbb{E}_{x\sim N}\Big[ (L^{\leq D}(x))^2 \Big]$

If $\chi^2_{\leq D}(P\|N) = o(1)$ , then $P$ and $N$ are "low-degree indistinguishable" at degree $D$ ; no test based on polynomials of degree at most $D$ can significantly differentiate them.

In information-theoretic contexts (e.g., symmetric distributions on $\{0,1\}^n$ ), low-degree indistinguishability means exact matching of $j$ -wise marginals for $j \leq k-1$ , and bounds are sought for the maximal statistical distance achievable by $k$ -wise marginal tests (Williamson, 2021).

2. Quantum Indistinguishability and Optical Measurement Schemes

In quantum optics, low-degree indistinguishability quantifies the genuine mutual indistinguishability among $n$ photons or particles, underpinning coherent quantum interference crucial for quantum computation (Pont et al., 2022). The parameter $\mathcal{I}_n$ ("genuine $n$ -photon indistinguishability") encodes the convex weight of the fully indistinguishable pure state component in an $n$ -photon mixed state:

$\rho = c_1 \rho^\parallel + \sum_{i>1} c_i \rho_i^\perp,\quad \mathcal{I}_n := c_1$

Cyclic multiport interferometry provides a direct measurement: the visibility $V_n$ of the induced $n$ -photon quantum fringe in the output coincidence probability equals $\mathcal{I}_n$ . Explicitly, the fringe oscillation amplitude in the $2n$-mode cyclic device is:

$P_{\text{ind}}(\alpha_1) \propto 1 + (-1)^n \cos\alpha_1$

$V_n = \frac{P_{\max} - P_{\min}}{P_{\max} + P_{\min}} = \mathcal{I}_n$

Such low-depth, low-loss interferometric platforms can efficiently certify high-order coherence for photonic quantum computing, and the protocol is scalable with constant optical depth (Pont et al., 2022).

3. Quantum Information, Phase Estimation, and Entropy of Mixing

Low-degree indistinguishability plays a critical role in quantum metrology and statistical mechanics. In interferometric phase estimation, the overlap parameter $\eta = |\langle\psi_a|\psi_b\rangle|^2$ between quantum probe states governs the quantum Fisher information (QFI):

$F_Q(\eta) = 2\eta + 2$

Even infinitesimal $\eta>0$ yields a quantum advantage in phase estimation over the shot-noise limit, and experimental results confirm this scaling in photonic setups (Knoll et al., 2019). In noisy regimes, a positive threshold of overlap is required to beat classical performance.

In statistical mechanics, the von Neumann overlap parameter $\lambda=|\langle\psi_1|\psi_2\rangle|$ provides a physically and mathematically continuous measure of indistinguishability, resolving the Gibbs paradox and enabling continuous entropy of mixing (Unnikrishnan, 2018):

$s(\lambda) = -k_B \left[ \tfrac{1+\lambda}{2} \ln \tfrac{1+\lambda}{2} + \tfrac{1-\lambda}{2} \ln \tfrac{1-\lambda}{2} \right]$

This continuity replaces the classical binary criterion (fully distinguishable vs. indistinguishable) with a quantum information-theoretic spectrum.

4. Algorithmic and Statistical Limits in Average-case Complexity

The "low-degree heuristic" in average-case complexity postulates that if all low-degree moments of planted and null models are sufficiently close, no efficient algorithm (or statistic based on low-degree polynomials) can distinguish between them (Hsieh et al., 9 Jan 2026). Recent rigorous results transform LDLR upper bounds into lower bounds on the power of noise-tolerant distinguishers:

For $\mathsf{P}$ over $\{0,1\}^n$ low-degree indistinguishable from uniform, any noisy version cannot be statistically distinguished from uniform.
For $\mathsf{P}$ over $\mathbb{R}^n$ low-degree indistinguishable from $\mathcal{N}(0,1)^n$ , no symmetric polynomial of degree $O(\log n/\log\log n)$ can distinguish noisy instances from true Gaussian.
For $\mathsf{P}$ over $\mathbb{R}^{n\times n}$ (Wigner matrix), no constant-sized subgraph count statistic can distinguish noisy versions from GOE.

Techniques leverage orthogonal polynomial expansions (Krawtchouk, Hermite), hypercontractivity, and analytic bounds to formalize these limitations.

5. Quantitative Indistinguishability in Combinatorics and Percolation

In percolation theory and graph processes, quantitative (low-degree) indistinguishability describes the number or structure of equivalence classes under factor-of-IID site percolations (Csóka et al., 8 Dec 2025). For any $\eta$ -non-hyperfinite clusters in FIID percolations on nonamenable Cayley graphs:

Properties (qI) and (qSI) (bounded number of indistinguishability classes) are equivalent to the "Sparse implies Thin" property (SiT): any such percolation must occupy density at least $c(G,\eta)>0$ .
In free groups (e.g., regular trees), explicit entropy bounds show $c(d,\eta)=\exp(-C_d/\eta)$ , while SiT can fail for non-exact groups.
In finite large-girth $d$ -regular graphs, FIID subgraphs of average degree at least $2+\delta$ have density lower-bounded by $c(d,\delta)>0$ .
Entropic inequalities, notably star-vertex entropy bounds, underlie these distinctions.

These combinatorial indistinguishability results are quantitative generalizations of ergodicity and mixing properties in infinite and finite graph processes.

6. Low-Degree Indistinguishability in Graph Theory and Homomorphism Counting

In structural graph theory, low-degree indistinguishability is formulated via homomorphism counts from classes of bounded-degree graphs (Roberson, 2022). For a family $\mathcal{F}$ , graphs $G,H$ are homomorphism indistinguishable if $\hom(F,G) = \hom(F,H)$ for all $F \in \mathcal{F}$ . Homomorphism indistinguishability over all graphs is equivalent to isomorphism, but restricting $\mathcal{F}$ changes the strength:

For the family $\mathcal{G}_\Delta$ of bounded-degree graphs, there exist non-isomorphic $H,H'$ such that $H \cong_{\mathcal{G}_\Delta} H'$ .
Weak oddomorphisms (parity-constrained homomorphisms) obstruct distinguishing certain graphs by low-degree test-graphs, and the closure properties of $\mathcal{G}_\Delta$ guarantee that homomorphism indistinguishability over bounded-degree graphs is strictly weaker than isomorphism.
Conjecturally, every family closed under minors and disjoint unions corresponds to a distinct homomorphism indistinguishability relation, and removal of any graph strictly coarsens the relation.

The parity-twist construction and oddomorphism machinery formalize these relations, with explicit counterexamples provided (e.g., $d$ -star $K_{1,d}$ and its parity-twisted versions).

7. Experimental Realizations and Practical Implications

The ability to probe and manipulate low-degree indistinguishability has direct experimental and practical impact:

In photonic systems, tuning spatial overlap or symmetry enables controlled activation of quantum-coherence resources (Sun et al., 2021).
Even small spatial overlaps (low $\mathcal{I}$ ) produce quantifiable $l_1$ -coherence, verified via precise density matrix tomography.
Phase-discrimination and estimation tasks benefit from indistinguishability-triggered quantum enhancement, observable as lower error probability than any incoherent strategy.
Bosonic and fermionic statistics can be simulated, enabling direct observation of how exchange symmetry alters coherence and metrological performance.

Recent experiments confirm that low-degree indistinguishability can be harnessed on demand for precision information processing, and may further inform the design of scalable photonic quantum computers.

Summary Table: Domains and Metrics of Low-Degree Indistinguishability

Domain	Metric / Technique	Key Quantitative Measure
Quantum optics	Cyclic interferometer; fringe	Fringe visibility $\mathcal{I}_n$
Complexity theory	LDLR, moment matching	$\chi^2_{\leq D}(P\\|N)$ , TV distance
Information theory	Marginal matching, polynomials	Maximum statistical distance $\Delta(n,k)$
Statistical mechanics	von Neumann overlap	Mixed entropy $s(\lambda)$
Percolation theory	Indistinguishability classes, entropy	Density lower bound $c(G,\eta)$
Graph theory	Homomorphism counts, oddomorphisms	Equality of $\hom(F,G)$ over family $\mathcal{F}$

Low-degree indistinguishability unifies algorithmic hardness, quantum resources, statistical limits, and structural graph properties by constraining what can be inferred by tests of bounded algebraic or combinatorial complexity. Its rigorous characterization is now central in quantum certification protocols, complexity-theoretic lower bounds, and structural combinatorics.