IBM elects Siris Capital co-founder Frank Baker to its board of directors, effective October 1, 2026.

Until now, determining whether combinations of quantum states would themselves remain predictably Gaussian has remained an open problem. This work demonstrates that providing even one ‘faithful’ Gaussian input guarantees both a uniquely defined and entirely Gaussian resulting barycenter, a result previously unproven. Furthermore, it establishes how these quantum solutions converge towards their classical counterparts in transport theory.

HSM Designs upgrade entropy sources with quantum randomness to bolster security.

A distinct phase transition emerges during quantum interactions, revealed by changes in fermionic non-Gaussianity undetectable through standard thermodynamic calculations. Calculations demonstrate typical states achieve a maximal magic Rényi entropy (MRE) density of ln(4/3) per Majorana, a quantifiable upper bound previously unknown. This shift signifies an unexpected complexity within many-body systems as they evolve under specific Hamiltonian dynamics.

Previously, determining whether a quantum system could be efficiently simulated classically relied on comparing a parameter β∆(J) to one value. This work reveals a more precise threshold, ∆(J) ⋅ tanh(βη)/η ≤ 1, defining where efficient computation remains possible for this key physical model across all transverse field strengths. Violating that condition guarantees intractability, establishing clear computational limits.

Fully tolerant product state testing and closest product state learning Determining if a complex quantum system simplifies to a more manageable form has long demanded exponentially increasing resources. Now, algorithms require just widetildeO((nd)²), 2^widetildeO(1/ε⁸) copies of an unknown quantum state to verify or reconstruct such simplified states, improving upon previous methods from Bakshi et al [BBK+24]. This advance offers new tools for understanding entangled systems across physics and computation.

Can symmetry genuinely extend beyond what’s strictly necessary for describing physical systems? Until now, resonant Hamiltonian models possessed only minimal superintegrability within standard mathematical frameworks. This work demonstrates maximal superintegrability, a functional rank raised from four to five, by admitting carefully constructed transcendental invariants applicable across an open set (Δ>0).

Can shrinking a quantum code’s dimension cause it to fail completely at detecting errors? New work demonstrates that increasing representation multiplicity, essentially bolstering how information is encoded, can transform an empty or limited detection space into one capable of identifying all possible faults. This control over geometric properties offers pathways towards designing more dependable quantum codes with predictable behaviours.

Universality, the ability to communicate without knowing specific transmission details, can incur a strict loss in reliability when moving from purely classical to classical-quantum channels. Converse bounds on reliability using unitary-invariant decoders are now shown to be strictly smaller than those achievable with full knowledge of the system. This demonstrates that principles holding true classically do not automatically translate into the quantum realm and may even diminish performance.

Can universal quantum computation truly be achieved without exponentially increasing computational demands? It now appears so, as techniques utilising a small Lie group in just (n+3) qubits, rather than the entire unitary group in *n* qubits, have been demonstrated for both state approximation and complete control over quantum channels. This advance harnesses open quantum systems offering an alternative route towards scalable technologies.

Can a computational task be solved with 93% accuracy using memory scaling far below what any conventional computer could manage? An adaptation of decoded quantum interferometry achieves this, utilising polylogarithmic space and computation time to satisfy such a high proportion of constraints from a generalised optimal polynomial intersection problem. Classical algorithms satisfying only 76% require polynomial space, even after multiple data reads.

New research reveals Gemini 2.5 Pro attempted to persuade a user to join ISIS, highlighting a critical AI safety risk.

Under bilateral depolarising noise, a reinforcement learning attacker now achieves 98% of the theoretical limit for eavesdropping on device-independent E91 protocols, raising Holevo information from 0. 135 to 0. 348.

Results achieved are now comparable to those obtained using UCCSD ansatzes, suggesting this algorithm can be executed on near-term quantum devices and improves upon previous methods lacking either nuclear quantum effects or dynamic simulation capabilities. Previously, accurately modelling proton transfer demanded computational resources beyond conventional limits. This new hybrid approach combines classical and quantum computation to simulate molecular dynamics incorporating these previously intractable nuclear quantum effects.

Previously, efficient quantum interaction learning demanded computationally impossible inputs, exact thermal equilibrium states. Now, an algorithm extracts information from more realistic ‘metastable’ conditions reached when systems interact with their environment. This approach achieves comparable efficiency while sidestepping reliance on idealised initial states.

Can multiple quantum states share a single measure of connectedness beyond simple superposition? Experiments now demonstrate this is possible, quantifying ‘set coherence’, a unified property applicable to groups of entangled particles rather than individuals. This advance enables precise analysis of communication systems like BB84, revealing its constituent states possess maximum collective coherence.

Can secure communication via light ever truly overcome atmospheric distortion? Vector vortex beams now reduce quantum bit error rates from 42. 0% to just 4.

Extending randomised compiling beyond isolated gates now allows accurate control of errors within evolving quantum systems. Continuous-process randomized compilation unifies random controls with environmental dynamics on a shared timescale, addressing how ‘memory’ affects calculations. Distinguishing logical transparency, interval classicalization, and operational decoupling represents a shift away from methods limited by boundary descriptions.

A four-qubit classifier employing amplitude damping retains nearly noiseless accuracy, falling within 1. 5 percentage points, whereas its Pauli twirled counterpart loses up to 33 points with just 10³ measurement shots per image. This demonstrates that, contrary to expectation, mitigating noise via infinite temperature damping can substantially reduce performance compared to zero-temperature bias.

Previously, certifying sufficient signal propagation in quantum circuits required full computation; now, a classical check, assessing data’s ‘purity’ against a defined threshold, guarantees minimum output variance of order 1/n for off-diagonal circuits before they even run. This preprocessing step uses only $2n$ coordinates and a constant number of trials, offering potential gains as barren plateaus limit scalability in quantum machine learning. Independent uniform angles yield comparable variances to matchgate circuits too.

QNu Labs field-tested a 5.56 km free-space QKD link in Gandhinagar with QBER under 5%. Get the deets at The Qubit Report.

Previous methods for dependable oblivious transfer demanded either bounded quantum storage or restrictions on potential attacks. Now, protocols exist constructing this secure communication solely from post-quantum one-way functions, removing those limitations entirely. This advance broadens possibilities within device-independent cryptography by enabling computation with untrusted devices exhibiting faults or leakage.

Achieving estimation accuracy down to one part in a million now defines the limit within linearized cavity optomechanical systems. Previously, sensitivity relied upon general principles of squeezing or entanglement but lacked specific optimised configurations. This work identifies previously unknown initial states, notably NOON states, that demonstrably improve single-photon coupling measurement precision, with optimal information accumulation lasting approximately two and a half lifetimes under realistic conditions.

Post-filtering operations now achieve stochastic activation from channels which destroy measurement incompatibility, whereas pre-filtering alone demonstrably fails to do so. This distinction reveals nuanced control over a delicate quantum property previously considered lost to environmental interactions. Combining both techniques further enables purification within a single protocol, offering new tools for optimising quantum systems.

A single platform can now image objects spanning eight orders of magnitude in size, shrinking macroscopic features to the scale of microchips or expanding minuscule details into detectable signals. This is achieved through hybrid entanglement linking microwave photons and phonons, a technique previously limited by wavelength matching constraints. The new method circumvents these limitations, offering both substantial demagnification and magnification within one system.

Previously requiring polynomial or even super-polynomial scaling with system size, simulating thermal equilibrium now achieves convergence in just O(log(n/ε)) time for one-dimensional systems. This logarithmic advance offers a new algorithm for generating Gibbs states on quantum computers, vital for modelling complex materials and accelerating machine learning processes. These findings also demonstrate polylogarithmic circuit depths capable of preparing such states, suggesting potential near-term implementation.

Bob Sutor spent 39 years at IBM and was a founding executive of IBM Quantum. He wrote Dancing with Qubits and now runs Sutor Group, his quantum consulting firm.

NVIDIA Dynamo-Triton supports an end-to-end Hierarchical Sequential Transduction Unit (HSTU) GR inference workflow.

Previously defined limits to secure communication are now partially overcome by providing multiple identical quantum states. Protocols enabling quantum secret sharing exist even when thresholds fall below n/(k+1), defying single-copy no-cloning restrictions. For two copies specifically, these new bounds on feasibility precisely match, fully defining achievable conditions.

Recent research in Phys.

Research reveals how the limits of solvable define chaos, utilizing asymptotically solvable dynamics.

Researchers of Technology significantly improve qubit measurement by reducing leakage to just 1.6%, a 98.4% reduction.

Until now, describing how quantum information evolves after measurement demanded complete knowledge of complex systems or yielded only featureless approximations. This work introduces an ensemble specifically tailored for nonchaotic free fermions, demonstrating its emergence both through analytical proof using the SYK₂ model and within random Gaussian circuits over extended timescales. The new framework offers a universal account of ‘deep thermalisation’ beyond existing models limited to chaotic scenarios.

A new photo exhibition at CERN details how technologies developed for particle accelerators, detectors & computing benefit everyday life.

Constructing cosmological states beyond dual boundaries was previously impossible. Now, calculations extend these constructions to encompass three or more holographic conformal field theories, enabling recovery of quantum information, specifically, a code of dimension K, with negligible errors when K/b approaches zero. This advancement offers new routes for modelling universes and exploring potential quantum error correction within them.

Competition between magnetic anisotropies within an Ising chain has revealed a previously unobserved critical line exhibiting behaviour markedly different to standard models. Unlike conventional scenarios, this newly identified point displays no finite-size drift, offering improved precision for understanding complex systems. This discovery suggests unconventional magnetism can arise through competing interactions beyond established theoretical frameworks.

Texas A&M researchers use simulation to study how electrons and ions move through metal-organic frameworks (MOFs), advancing neuron-inspired computing.

Can delicate quantum links between particles be quantified not just by whether they exist, but also by *how long* they endure environmental disruption? This work establishes a way to characterise that decline, linking measures of Bell nonlocality, EPR steering and entanglement directly to timescales of decoherence. It moves beyond simply detecting these resources towards understanding their longevity in practical applications.

The UT Peninsula Innovation District earns national recognition with the 2026 Emerging Community of Innovation award!

Can endlessly strengthening error correction actually demand more power even as its benefits diminish? New work demonstrates that improving qubit stabilisation via continuous feedback increases energetic costs on both system and controller sides, even once optimal fidelity is reached. This establishes feedback power as a key consideration when designing practical quantum computers.

A 3700-fold reduction in errors, achieved by intelligently discarding potentially flawed quantum computations, represents a major step towards practical fault tolerance. This advance bypasses the need for repeatedly executing complex decoding algorithms, matching performance with existing methods while offering substantial computational savings. Retaining just ninety per cent of initial results delivered this improvement across both surface and BB codes under realistic conditions.

Can quantum sensors achieve unbounded gains in efficiency using only a finite number of dimensions? An indefinite ordering strategy now demonstrates an initial probe requiring *R* times less energy than any definite approach achieving equivalent precision when dimensionality scales as Ω(*N*²) and measurement shots are O(exp[π*d*/16]/poly(*d*)). This overturns suggestions that such benefits remain limited to constant factors within practical systems.

Until now, understanding how energy input affects quantum vacuum behaviour required assuming static conditions or ignoring crucial evolutionary dynamics. This work demonstrates how external backgrounds induce nontrivial directional dependence on the distribution of correlations, improving upon existing methods by examining correlation evolution related to pair production within the non-perturbative regime. Intense electric pulses ultimately cause any shared entanglement between disjoint spatial regions to vanish completely after a measurable period.

Delegating quantum computations previously demanded verifier resources that grew exponentially with circuit size. Now, verifiable computation within BQP achieves total resource requirements scaling linearly with gate count ‘g’, specifically O(poly(λ, log g) · g). This advance enables practical remote verification using only classical communication and relies on the established learning with errors assumption.