April 2025

We present the first implementation of text-level quantum natural language processing, a field where quantum computing and AI have found a fruitful intersection. We focus on the QDisCoCirc model, which is underpinned by a compositional approach to rendering AI interpretable: the behaviour of the whole can be understood in terms of the behaviour of parts, and the way they are put together. Interpretability is crucial for understanding the unwanted behaviours of AI. By leveraging the compositional structure in the model’s architecture, we introduce a novel setup which enables ‘compositional generalisation’: we classically train components which are then composed to generate larger test instances, the evaluation of which asymptotically requires a quantum computer. Another key advantage of our approach is that it bypasses the trainability challenges arising in quantum machine learning. The main task that we consider is the model-native task of question-answering, and we handcraft toy scale data that serves as a proving ground. We demonstrate an experiment on Quantinuum’s H1-1 trapped-ion quantum processor, which constitutes the first proof of concept implementation of scalable compositional QNLP. We also provide resource estimates for classically simulating the model. The compositional structure allows us to inspect and interpret the word embeddings the model learns for each word, as well as the way in which they interact. This improves our understanding of how it tackles the question-answering task. As an initial comparison with classical baselines, we considered transformer and LSTM models, as well as GPT-4, none of which succeeded at compositional generalisation.

We present a protocol that maximizes unconditional entanglement generation between two masses interacting directly through $1/r^{n}$ potential. The protocol combines optimal quantum control of continuously measured masses with their non-equilibrium dynamics, driven by a time-dependent interaction strength. Applied to a pair of optically trapped sub-micron particles coupled via electrostatic interaction, our protocol enables unconditional entanglement generation at the fundamental limit of the conditional state and with an order of magnitude smaller interaction between the masses compared to the existing steady-state approaches.

We develop a feedback strategy based on optimal quantum feedback control for Gaussian systems to maximise the likelihood of steady-state entanglement detection between two directly interacting masses. We employ linear quadratic Gaussian (LQG) control to engineer the phase space dynamics of the two masses and propose Einstein-Podolsky-Rosen (EPR)-type variance minimisation constraints for the feedback to facilitate unconditional entanglement generation. This scheme allows for stationary entanglement in parameter regimes where strategies based on total energy minimisation ($cooling$) would fail. This feedback strategy, applied to the system of two masses driven out-of-thermal equilibrium [arXiv:2408.06251] enables unconditional entanglement generation under realistic experimental conditions.

Quantum computing and AI have found a fruitful intersection in the field of natural language processing. We focus on the recently proposed DisCoCirc framework for natural language, and propose a quantum adaptation, QDisCoCirc. This is motivated by a compositional approach to rendering AI interpretable: the behavior of the whole can be understood in terms of the behavior of parts, and the way they are put together. For the model-native primitive operation of text similarity, we derive quantum algorithms for fault-tolerant quantum computers to solve the task of question-answering within QDisCoCirc, and show that this is BQP-hard; note that we do not consider the complexity of question-answering in other natural language processing models. Assuming widely-held conjectures, implementing the proposed model classically would require super-polynomial resources. Therefore, it could provide a meaningful demonstration of the power of practical quantum processors. The model construction builds on previous work in compositional quantum natural language processing. Word embeddings are encoded as parameterized quantum circuits, and compositionality here means that the quantum circuits compose according to the linguistic structure of the text. We outline a method for evaluating the model on near-term quantum processors, and elsewhere we report on a recent implementation of this on quantum hardware. In addition, we adapt a quantum algorithm for the closest vector problem to obtain a Grover-like speedup in the fault-tolerant regime for our model. This provides an unconditional quadratic speedup over any classical algorithm in certain circumstances, which we will verify empirically in future work.

Recently, machine learning had a remarkable impact, from scientific to everyday-life applications. However, complex tasks often imply unfeasible energy and computational power consumption. Quantum computation might lower such requirements, although it is unclear whether enhancements are reachable by current technologies. Here, we demonstrate a kernel method on a photonic integrated processor to perform a binary classification. We show that our protocol outperforms state-of-the-art kernel methods including gaussian and neural tangent kernels, exploiting quantum interference, and brings a smaller improvement also by single photon coherence. Our scheme does not require entangling gates and can modify the system dimension through additional modes and injected photons. This result opens to more efficient algorithms and to formulating tasks where quantum effects improve standard methods.

Are you, with your perceptions, memories and observational data, a Boltzmann brain, namely a fleeting statistical fluctuation out of the thermal equilibrium of the universe? Arguments are given in the literature claiming that this bizarre hypothesis needs to be considered seriously, that all of our data about the past is actually a mirage. We point to a difficulty in these arguments. They are based on the dynamical laws and on statistical arguments, but they disregard the fact that we infer the dynamical laws presupposing the reliability of our data records about the past. Hence the reasoning in favor of the Boltzmann brain hypothesis contradicts itself, relying on the reliability of our data about the past to conclude that that data is wrong. More broadly, it is based on incomplete evidence. Incomplete evidence notoriously leads to false conclusions.

In this work, we demonstrate that quantizing gravity on a null hypersurface leads to the emergence of a CFT associated with each null ray. This result stems from the ultralocal nature of null physics and is derived through a canonical analysis of the Raychaudhuri equation, interpreted as a constraint generating null time reparametrizations. The CFT exhibits a non-zero central charge, providing a mechanism for the quantum emergence of time in gravitational systems and an associated choice of vacuum state. Our analysis reveals that the central charge quantifies the degrees of freedom along each null ray. Throughout our investigation, the area element of a cut plays a crucial role, necessitating its treatment as a quantum operator due to its dynamic nature in phase space or because of quantum backreaction. Furthermore, we show that the total central charge diverges in a perturbative analysis due to the infinite number of null generators. This divergence is resolved if there is a discrete spectrum for the area form operator. We introduce the concept of `embadons’ to denote these localized geometric units of area, the fundamental building blocks of geometry at a mesoscopic quantum gravity scale.

We illustrate Heisenberg’s method of matrix mechanics using the planar quantum rotor example. We show how to find the spectrum of this simple model without the need to use the eigenstates of the system. This then leads us to speculate on the role the Heisenberg state plays in quantum mechanics and to ask whether one could even completely dispose of the need for states.

The ZX-calculus is a graphical language for reasoning about quantum computing and quantum information theory. As a complete graphical language, it incorporates a set of axioms rich enough to derive any equation of the underlying formalism. While completeness of the ZX-calculus has been established for qubits and the Clifford fragment of prime-dimensional qudits, universal completeness beyond two-level systems has remained unproven until now. In this paper, we present a proof establishing the completeness of finite-dimensional ZX-calculus, incorporating only the mixed-dimensional Z-spider and the qudit X-spider as generators. Our approach builds on the completeness of another graphical language, the finite-dimensional ZW-calculus, with direct translations between these two calculi. By proving its completeness, we lay a solid foundation for the ZX-calculus as a versatile tool not only for quantum computation but also for various fields within finite-dimensional quantum theory.

We explore the notion of events at the intersection between quantum physics and gravity, inspired by recent research on superpositions of semiclassical spacetimes. By going through various experiments and thought experiments — from a decaying atom, to the double-slit experiment, to the quantum switch — we analyse which properties can and cannot be used to define events in such non-classical contexts. Our findings suggest an operational, context-dependent definition of events which emphasises that their properties can be accessed without destroying or altering observed phenomena. We discuss the implications of this understanding of events for indefinite causal order as well as the non-absoluteness of events in the Wigner’s friend thought experiment. These findings provide a first step for developing a notion of event in quantum spacetime.