No-Random-AQIM Theorem: Quantum Information Limits

• The paper shows that quantum measurement randomness cannot be certified beyond a formal-system-dependent finite bound.
• It demonstrates that no efficient randomness test can distinguish true quantum randomness from outputs of cryptographically secure pseudorandom generators.
• It reveals that Haar-random bipartite isometries almost never achieve approximate quantum masking, underscoring fundamental limits in subsystem architectures.

“No-Random-AQIM Theorem” is not a single canonical theorem title in the arXiv literature. In the provided corpus, it functions as an Editor’s term for several distinct no-go or no-certification results at the interface of quantum information, randomness, and algorithmic or masking-based formalisms. Its most literal use appears in approximate quantum information masking, where a bipartite Haar-random isometry almost never realizes good approximate masking (Li et al., 25 Jul 2025). The same label is also suggested for metamathematical limits on proving quantum-measurement randomness, for cryptographic indistinguishability of efficiently simulatable quantum randomness from pseudo-randomness under efficiently computable tests, for the absence of bound randomness in quantum nonlocality when randomness is generated from all inputs, and for uselessness theorems in randomized oracle models (Rogers, 2010).

1. Terminological status and conceptual range

The expression “AQIM” is itself context-dependent in the supplied literature. In approximate masking, it denotes approximate quantum information masking. In the algorithmic-randomness papers, it is suggested as a natural abbreviation for Algorithmic Quantum Information / Measurement, and the “No-Random-AQIM” label is used informally to capture the relevant no-go content rather than as an official theorem name (Tsurumaru et al., 2023).

Context	Central object	Core claim
Algorithmic randomness of measurement outcomes	$C(x\mid l(x))$, $0$-randomness	High randomness of specific quantum outputs cannot be proved beyond a finite bound
Efficient randomness testing of QRNGs	ECS QRNGs, CPRNGs, efficient $I$	No efficiently computable randomness measure distinguishes them beyond negligible advantage
Approximate quantum information masking	Random isometries $V:H_A\to H_{B_1\cdots B_m}$	Bipartite random isometries almost never realize AQIM
Device-independent randomness	Average guessing probability $P_g^{(q)}$	No bound randomness exists for quantum nonlocal behaviors when all inputs are used
Oracle models with internal randomness	Query uselessness	$k$ quantum queries are useless iff $2k$ classical pairwise queries are useless

These results are not interchangeable. Some concern provability of randomness, some concern efficient distinguishability, some concern random constructions of maskers, some concern device-independent certification, and some concern query complexity with internal random coins. A plausible implication is that “No-Random-AQIM Theorem” is best treated as a family resemblance term rather than a single theorem schema.

2. Unprovability of algorithmic randomness in quantum measurement

In the metamathematical setting of “Quantum Measurements Cannot be Proved to be Random” (Rogers, 2010), “random” means algorithmically random in the Kolmogorov–Martin-Löf sense. For a finite binary string $x=x_1x_2\ldots x_n$, with $l(x)=n$, the paper uses the conditional Kolmogorov complexity

$C(x\mid l(x))=\min_{U(l(x),A)=x} l(A),$

and the Martin–Löf reference universal test

$0$0

A string is $0$1-random when $0$2, and it is $0$3-random iff

$0$4

The theorem is a Chaitin-style incompleteness result. For any fixed effective formal system and universal machine, there exists a natural number $0$5 such that there is no proof of any statement of the form

$0$6

for any finite string $0$7. In the quantum-measurement interpretation, if a quantum random number generator outputs a bit string $0$8, then once $0$9 is large enough, the formal system cannot prove that

$I$0

that is, cannot prove that the observed output is $I$1-random.

The proof searches mechanically for a proof $I$2 that some $I$3 satisfies $I$4. If such a proof existed, a program $I$5 taking $I$6 as input could enumerate proofs and output $I$7, yielding an upper bound

$I$8

For sufficiently large $I$9, $V:H_A\to H_{B_1\cdots B_m}$0, contradicting the purported theorem that $V:H_A\to H_{B_1\cdots B_m}$1.

The result is explicitly relative to the chosen formal system and universal machine. It does not show that quantum measurements are not random, does not refute the Born rule, and does not establish hidden variables. It shows only that a fixed effective axiomatic framework cannot certify arbitrarily high algorithmic complexity for specific observed measurement strings. The same logic extends from finite prefixes to the impossibility of proving Martin-Löf randomness of a particular infinite measurement sequence.

3. Efficient indistinguishability of quantum randomness and pseudo-randomness

A different “No-Random-AQIM” reading appears in “Indistinguishability between quantum randomness and pseudo-randomness under efficiently calculable randomness measures” (Tsurumaru et al., 2023). Here the central assumptions are the existence of a cryptographic pseudo-random number generator and the efficient classical simulability of the relevant quantum random number generator. A randomness measure is formalized as an efficient probabilistic algorithm

$V:H_A\to H_{B_1\cdots B_m}$2

and its operational use is mediated by an efficient distinguisher $V:H_A\to H_{B_1\cdots B_m}$3.

The paper’s first no-go theorem states that if a CPRNG $V:H_A\to H_{B_1\cdots B_m}$4 exists, then for any efficient randomness measure $V:H_A\to H_{B_1\cdots B_m}$5 and any associated distinguisher $V:H_A\to H_{B_1\cdots B_m}$6,

$V:H_A\to H_{B_1\cdots B_m}$7

Thus no efficiently computable randomness measure can significantly distinguish true uniform outputs from CPRNG outputs.

The more general theorem replaces ideal uniform randomness by an efficiently classically simulatable physical system $V:H_A\to H_{B_1\cdots B_m}$8. If $V:H_A\to H_{B_1\cdots B_m}$9 is ECS, then for any polynomial $P_g^{(q)}$0 there exists a deterministic algorithm

$P_g^{(q)}$1

such that, for any efficient randomness measure $P_g^{(q)}$2 and distinguisher $P_g^{(q)}$3,

$P_g^{(q)}$4

In the paper’s own interpretation, this means that no efficiently computable AQIM-like measure can certify that outputs from an ECS QRNG are “more random” than outputs from an appropriately constructed pseudo-random source.

The restriction to efficient measures is essential. Lempel–Ziv complexity, normalized LZ complexity, Borel normality, and statistical test suites fall within scope because they are polynomial-time computable. Exact Kolmogorov complexity and Martin–Löf randomness do not. The ECS assumption is equally essential: without an efficient classical simulator, the reduction to CPRNG security does not go through.

The empirical sections are presented as consistency checks. On data from IBM Quantum coin tosses and the Innsbruck Bell experiment, LZ complexity and Borel normality are reported as essentially indistinguishable from those of carefully matched pseudo-random strings once bias and sample length are controlled. The paper therefore reinterprets earlier claims about algorithmic distinctions between quantum and pseudo-random outputs as effects of finite size, bias, or experimental imperfection rather than evidence of a fundamental difference.

4. Approximate quantum information masking and the literal no-random-AQIM theorem

The most literal occurrence of the phrase is in “Random approximate quantum information masking” (Li et al., 25 Jul 2025). In this setting, AQIM means approximate quantum information masking. Let

$P_g^{(q)}$5

be an isometry from a logical Hilbert space to a multipartite physical space, and let $P_g^{(q)}$6 be the image of a set of pure input states. Exact bipartite masking requires the reduced state on each local subsystem to be independent of the encoded state. The multipartite generalization is $P_g^{(q)}$7-uniform masking, meaning that for every subsystem $P_g^{(q)}$8 with $P_g^{(q)}$9,

$k$0

Approximate masking is quantified by figures of merit based on trace distance. For a subsystem $k$1, the maximum variation is

$k$2

and the maximum inaccuracy relative to the average code projector is

$k$3

The paper proves equivalences up to constant factors, such as

$k$4

so either quantity can serve as the AQIM error parameter.

The central negative result concerns bipartite Haar-random isometries. For a random $k$5-dimensional subspace $k$6, the expected variation obeys

$k$7

A concentration bound then gives, for any $k$8,

$k$9

From this, the paper concludes that the probability that a Haar-random bipartite isometry is a $2k$0-approximate masker with $2k$1 is exponentially small in $2k$2. This is the theorem explicitly described as a generalization of the original no-masking theorem to a no-random-AQIM theorem for bipartite systems.

The significance is architectural rather than metamathematical. Randomness in the isometry does not rescue bipartite masking: almost all random bipartite embeddings fail to make all codeword marginals nearly identical on both parties. The negative result is therefore a statement about the geometry of random subspaces, not about randomness certification or proof theory.

The multipartite case is the opposite. For $2k$3-partite systems with local dimension $2k$4, random subspaces typically satisfy

$2k$5

with high probability, and the failure probability is exponentially small in $2k$6. Corollaries in the paper state that the number of physical qubits required to randomly mask one logical qubit scales linearly in the number of logical qubits. The same formalism is then connected to approximate quantum error correction through inequalities relating masking inaccuracy $2k$7 and QEC inaccuracy $2k$8, leading to AQECCs with constant code rates and exponentially small correction inaccuracies.

5. No bound randomness in quantum nonlocality

“No Bound Randomness in Quantum Nonlocality” establishes a different theoremic pattern (Ramanathan et al., 10 Sep 2025). Rather than showing that randomness cannot be certified, it shows that bound randomness does not exist for quantum nonlocal behaviors once randomness is generated from all input pairs. For a bipartite quantum behavior $2k$9 and an input distribution $x=x_1x_2\ldots x_n$0 with full support, the average guessing probability against a quantum adversary is

$x=x_1x_2\ldots x_n$1

and the corresponding average min-entropy is

$x=x_1x_2\ldots x_n$2

The main theorem states that any quantum nonlocal behavior is a sufficient resource for device-independent randomness amplification of an $x=x_1x_2\ldots x_n$3-Santha–Vazirani source for any $x=x_1x_2\ldots x_n$4, in the sense that

$x=x_1x_2\ldots x_n$5

Equivalently, every nonlocal quantum behavior yields strictly positive average conditional min-entropy when all inputs are used for randomness generation.

The conceptual theorem in the appendix states more sharply that no nonlocal quantum behavior exhibits bound randomness against a quantum adversary. If Eve could always guess outputs perfectly once told the inputs, then the induced tripartite quantum behavior would imply a joint probability distribution over all settings, hence locality. Perfect predictability for all inputs therefore collapses nonlocality.

This must be distinguished from fixed-input randomness certification. The same paper recalls earlier results showing that there exist maximally nonlocal quantum behaviors for which

$x=x_1x_2\ldots x_n$6

for every fixed input pair $x=x_1x_2\ldots x_n$7. Such behaviors are useless for spot-checking protocols that generate randomness from a single setting, yet they still satisfy

$x=x_1x_2\ldots x_n$8

for any full-support input distribution. The contrast is operational: fixed-setting guessing probability is not a faithful nonlocality measure, whereas the average guessing probability is shown to be faithful and monotonic under WCCPI operations.

The paper also applies this viewpoint to detection efficiency. For the $x=x_1x_2\ldots x_n$9 inequality, the average min-entropy is strictly positive for any violation $l(x)=n$0, and the threshold detection efficiency for certifying nonzero average min-entropy coincides with the threshold for observing nonlocality. Analytic formulas are additionally derived for the average guessing probability of a single party’s output in the CHSH scenario, with recovery of the Pironio et al. fixed-input expression as the special case $l(x)=n$1.

6. Internal randomness in oracle models and uselessness theorems

A further, structurally different no-go appears in “Uselessness for an Oracle Model with Internal Randomness” (Harrow et al., 2011). Here the oracle applies a permutation selected according to internal random coins. A single query is the CPTP map

$l(x)=n$2

where

$l(x)=n$3

The relevant notion is uselessness, meaning that a bounded number of queries yields no advantage over prior guessing. In the randomized setting, the key classical notion is pairwise classical uselessness: $l(x)=n$4 classical queries are arranged in $l(x)=n$5 pairs sharing the same internal seed. The paper’s central theorem states that

$l(x)=n$6

In the deterministic special case, this reduces to the statement that $l(x)=n$7 quantum queries are useless iff $l(x)=n$8 classical queries are useless, extending the earlier Meyer–Pommersheim result.

The proof uses an oracle-state encoding. For deterministic permutation oracles, one defines

$l(x)=n$9

and in the randomized case the corresponding mixed encoding is

$C(x\mid l(x))=\min_{U(l(x),A)=x} l(A),$0

One query can prepare one encoding state, and conversely one encoding state can simulate one query with heralded success probability $C(x\mid l(x))=\min_{U(l(x),A)=x} l(A),$1. Quantum uselessness is then equivalent to equality of the class-conditional average encodings, while pairwise classical transcripts determine the relevant matrix elements of those encodings.

This theorem is not about randomness certification in the quantum-foundational sense. It is about the inability of quantum or classical query algorithms to extract any nonzero information from an oracle family with internal randomness. Even so, it fits the broader “No-Random-AQIM” pattern in the supplied material because the internal random coins do not create an unbounded advantage regime beyond the precise $C(x\mid l(x))=\min_{U(l(x),A)=x} l(A),$2-quantum versus $C(x\mid l(x))=\min_{U(l(x),A)=x} l(A),$3-classical correspondence.

7. Comparative interpretation, scope, and common confusions

The supplied literature supports five non-equivalent readings of “No-Random-AQIM Theorem.” First, it can mean a metamathematical incompleteness statement: specific quantum-measurement strings cannot be proved algorithmically random beyond a formal-system-dependent bound (Rogers, 2010). Second, it can mean an efficient indistinguishability theorem: no polynomial-time randomness measure separates ECS quantum outputs from CPRNG outputs under standard cryptographic assumptions (Tsurumaru et al., 2023). Third, it can denote the literal masking theorem: almost all random bipartite isometries fail to realize approximate quantum information masking (Li et al., 25 Jul 2025). Fourth, it can refer to the absence of bound randomness in quantum nonlocality when average guessing over all inputs is used (Ramanathan et al., 10 Sep 2025). Fifth, it can describe uselessness of randomized oracle access in the unbounded-error query model (Harrow et al., 2011).

Several misconceptions are addressed directly by these papers. The incompleteness theorem does not show that quantum mechanics lacks randomness. The efficient-indistinguishability theorem does not apply to noncomputable tests or to quantum processes that are not efficiently classically simulatable. The masking theorem is negative only in the bipartite random-isometry regime and becomes positive in multipartite systems. The nonlocality theorem excludes bound randomness only against quantum adversaries and with full-support input use; no-signalling adversaries behave differently. The oracle uselessness theorem concerns advantage over prior guessing, not the truth of physical randomness claims.

A plausible synthesis is that the common theme is not “randomness is absent,” but that claims about randomness are sharply constrained by the surrounding framework: by the proof system in algorithmic information theory, by computational efficiency in cryptographic distinguishability, by subsystem architecture in masking, by protocol structure in device-independent randomness generation, and by query access models in oracle complexity. Under that interpretation, “No-Random-AQIM Theorem” designates a family of precise limitations on how randomness can be certified, realized, or exploited in quantum-information settings rather than a single universally accepted theorem name.