Quantum Functional Information (QFI)

• Quantum Functional Information (QFI) is defined as -log₂(P(f)), quantifying how rare a high-fidelity quantum state is within a circuit ensemble.
• Empirical studies reveal that near-perfect yet uncommon states yield higher QFI than frequently attained exact states, promoting diversity.
• QFI-driven optimization encourages exploration by preserving circuit diversity and robustness, distinctly differing from fidelity-only and Quantum Fisher Information measures.

Quantum Functional Information (QFI) is a rarity-aware measure of how informative a quantum state or circuit is, defined not only by task performance but also by how uncommon that performance is within the search space of possible quantum configurations. In the formulation introduced in "Quantum Functional Information through the Evolution of Random Circuits" (Pasti et al., 14 Sep 2025), QFI extends the logic of classical functional information to Hilbert space: a configuration is more informative if it is hard to find by random search yet still achieves a desired function. The resulting quantity is explicitly designed to capture the balance between functionality and rarity within the Hilbert space, and it is applied to both randomly sampled quantum circuits and evolutionary search over circuit families.

1. Definition and mathematical formulation

The starting point is classical functional information,

$I(F) = -\log_2 \Pr[f(c)\ge T],$

where $f(c)$ is a functional score for configuration $c$, and $T$ is a threshold defining success. In the quantum extension, the same definition is applied to circuits or states in Hilbert space. For a circuit $c\in C$, let

$f(c): C \to \mathbb{R}$

be a functional evaluation, such as fidelity to a target state. Then QFI for threshold-based functionality is defined as

$I(F) = -\log_2 \Pr[f(c)\ge T].$

When the focus is a specific fidelity value rather than a threshold, the corresponding functional-information value is written as

$I(f) = -\log_2 P(f),$

where $P(f)$ is the empirical probability of observing fidelity $f$ in the circuit ensemble (Pasti et al., 14 Sep 2025).

For a pure target state $f(c)$0, fidelity is computed as

$f(c)$1

and if both are pure states, this becomes

$f(c)$2

Within this empirical setting, Quantum Functional Information is therefore the negative base-2 logarithm of the probability of observing a given fidelity level. A low-probability fidelity outcome corresponds to high QFI. This construction makes the central quantity a log-probability measure of rarity conditioned on functionality, rather than a direct state-distance or fluctuation measure (Pasti et al., 14 Sep 2025).

The formalism is explicitly probabilistic: the magnitude of QFI is determined by how often functionally similar circuits appear in the ensemble under study. This suggests that QFI is inseparable from the circuit distribution used to estimate $f(c)$3, and that its interpretation is intrinsically ensemble-relative rather than universal in the metrological sense.

2. Distinction from fidelity and from Quantum Fisher Information

Quantum Functional Information is not the same as fidelity, entropy, or entanglement measures. Fidelity measures closeness to a target state and answers the question “How accurate is this state?” Entropy measures mixedness or local uncertainty and answers the question “How disordered or entangled is this state?” Quantum Functional Information measures how rare it is to obtain a functionally good state by chance and answers the question “How informative is this state because it is both useful and uncommon?” (Pasti et al., 14 Sep 2025).

A separate source of ambiguity is acronymic. In quantum metrology, QFI usually denotes Quantum Fisher Information, the quantity that quantifies the sensitivity of a quantum state to infinitesimal parameter changes and determines the ultimate precision of parameter estimation through the quantum Cramér–Rao bound (Beckey et al., 2020). Quantum Fisher Information is defined, for example, through fidelity-based or symmetric-logarithmic-derivative formulations, and it is a standard metrological resource (Sone et al., 2020).

Quantity	Defining question	Representative expression
Quantum Functional Information	How informative is a state because it is both useful and uncommon?	$f(c)$4
Fidelity	How accurate is this state?	$f(c)$5
Quantum Fisher Information	How sensitively does a state change under parameter variation?	$f(c)$6

This distinction is substantive rather than terminological. Quantum Fisher Information is tied to parameter-estimation sensitivity, Bures geometry, and the quantum Cramér–Rao bound (Beckey et al., 2020), whereas Quantum Functional Information is tied to empirical rarity under a circuit ensemble and to the distribution of functional performance levels (Pasti et al., 14 Sep 2025). A common misconception is that these two notions are variants of the same measure; the available formulations do not support that identification.

3. Random-circuit estimation and empirical behavior

The empirical validation of Quantum Functional Information is based on random circuit sampling. The study randomly generated 5 million quantum circuits for $f(c)$7 qubits, with circuit size up to 50 gates. The gate set included

$f(c)$8

For each circuit, fidelity was computed against the Bell state

$f(c)$9

for 2 qubits, and against the GHZ state

$c$0

for $c$1 (Pasti et al., 14 Sep 2025).

The estimation pipeline is specified as follows:

1. Randomly sample circuits.
2. Compute fidelity of each circuit output with the target state.
3. Bin fidelities into 200 intervals.
4. Estimate the probability distribution $c$2 via regression.
5. Convert probability to QFI using

$c$3

The fidelity distribution was modeled with decision-tree regression, and the resulting QFI curve was smoothed by spline-based regression with ridge regularization (Pasti et al., 14 Sep 2025).

Several empirical regularities are reported. Most random circuits have low fidelity, and high-fidelity states are rare. As qubit number increases, functional states become rarer, so QFI increases with system size. A particularly important observation is that states with fidelity exactly $c$4 often have lower QFI than states with fidelity near $c$5, such as $c$6 or $c$7. In this sense, exact target states are not always the most informative in the QFI sense, because they may be more frequently attained than narrowly suboptimal states (Pasti et al., 14 Sep 2025).

The same random-sampling study reports a negative Pearson correlation between fidelity and circuit complexity: fidelity versus number of gates is about $c$8, and fidelity versus depth is about $c$9. This indicates that high-fidelity circuits tend to be shallower and use fewer gates. The paper does not infer from this that complexity is irrelevant; rather, QFI can still favor certain complex circuits if they are rarer and functionally meaningful (Pasti et al., 14 Sep 2025).

4. Evolutionary search and optimization behavior

Quantum Functional Information is also used as an objective in evolutionary circuit search. The evolutionary algorithm starts from a random population of circuits, each encoded as a gate sequence and evaluated on several metrics: fidelity, average single-qubit entropy, robustness, depth, and number of gates. Scoring is performed using either a fidelity objective or a QFI objective; the top 40% are kept as elites, and mutation is implemented by gate replacement, insertion, deletion, and angle perturbation (Pasti et al., 14 Sep 2025).

When fidelity is the optimization target, the algorithm rapidly converges to $T$0. This convergence is accompanied by a narrow fidelity distribution, small interquartile range, collapse of population diversity, and trapping in a small region of the solution space. The reported interpretation is that fidelity optimization is efficient at exploitation but poor at exploration (Pasti et al., 14 Sep 2025).

When QFI is the objective, the fidelity distribution remains much broader. The reported features are a larger interquartile range, more intermediate and high-fidelity circuits, greater structural diversity, and less collapse into one dominant solution. The interquartile range more than doubles under QFI-driven evolution compared with fidelity-driven evolution. Mean fidelity is lower, but the population explores more of the landscape (Pasti et al., 14 Sep 2025).

The highest QFI values occur not at $T$1, but around high yet imperfect fidelities such as $T$2–$T$3. QFI-driven evolution also tends to produce circuits that are more robust under noise, lower in entropy, slightly deeper, with more gates, and more structured. This suggests that QFI functions as an exploration-preserving objective that rewards rare high-performing regions rather than immediate convergence to common solutions.

5. Interpretation, scope, and common misconceptions

The central interpretive claim of the framework is that QFI captures functional specificity in Hilbert space. Many states are possible, but only a tiny subset are useful for a given purpose—for example, Bell or GHZ states for entanglement-based tasks. QFI is intended to quantify how informative such states or circuits are by combining usefulness with scarcity (Pasti et al., 14 Sep 2025).

One recurrent misconception is that a state with fidelity $T$4 must automatically be the most informative. The reported random-circuit results explicitly contradict this: near-perfect but common states can have lower QFI, while slightly suboptimal but rarer states can have higher QFI. The distinction is therefore between common success and rare near-success, not between success and failure alone (Pasti et al., 14 Sep 2025).

A second misconception is that QFI is just another accuracy score. The formal definition does not support that interpretation. Fidelity is an input to the empirical construction, but QFI is computed from the probability of observing a fidelity level, not from the fidelity level alone. The same fidelity value can therefore correspond to different informational significance under different ensemble frequencies. This suggests that QFI is a property of performance-in-context rather than performance in isolation.

A third misconception arises from the acronym overlap with Quantum Fisher Information. Quantum Fisher Information is a metrological quantity governing parameter-estimation precision, often introduced through the quantum Cramér–Rao bound and fidelity or SLD formulas (Beckey et al., 2020); Quantum Functional Information is a log-probability measure of rarity conditioned on functionality (Pasti et al., 14 Sep 2025). The two measures answer different questions, rely on different mathematical objects, and occupy different research programs.

6. Applications and research context

The proposed uses of Quantum Functional Information are fourfold. In circuit design, it can guide the search for circuits that are not merely high-fidelity, but also rare and structurally meaningful, helping avoid overfitting to easy solutions. In benchmarking, because it reflects how difficult a circuit is to obtain by random generation, it can serve as a benchmark for the intrinsic complexity of target states or task families. In variational quantum algorithms, it can be used as a cost function favoring solutions that are robust and representative of rare functional regions rather than just maximizing a narrow performance score. In emergent pattern discovery, it may help identify unexpected or emergent structures in quantum state spaces by highlighting states that are functionally special but statistically uncommon (Pasti et al., 14 Sep 2025).

These applications place Quantum Functional Information alongside a broader family of resource-oriented quantum metrics, but not within the metrological role occupied by Quantum Fisher Information. Quantum Fisher Information remains the standard quantity for estimating unknown parameters, certifying metrological sensitivity, and relating local state distinguishability to the Bures metric (Sone et al., 2020). Quantum Functional Information, by contrast, is directed at ranking states and circuits according to usefulness plus scarcity within a chosen ensemble (Pasti et al., 14 Sep 2025).

The present formulation is therefore best understood as an ensemble-based design and analysis tool for quantum circuits and target-state preparation. Its defining conceptual contribution is the claim that a state that is slightly less perfect but much rarer may be more informative than a common perfect state. Within the framework that has been reported, this is the central criterion by which Quantum Functional Information differs from fidelity-only evaluation.