arXiv subjects

Qinyuan Zheng

Publications and source records attributed to Qinyuan Zheng.

11 recordsLinked to original sources

arXiv2026

How Good Are Frontier Models at Physics? Expert Re-Grading Reveals Broken Evaluations and Near-Saturation of Leading Benchmarks

Low reported scores on leading physics benchmarks, including those featured in the Artificial Analysis Intelligence Index (2026), suggest that frontier language models still struggle with advanced physics, a demanding test of their scientific reasoning and quantitative problem-solving abilities. Yet this impression does not always align with domain experts' experiences using these models in their work. We revisit these reported findings by evaluating frontier models on six widely used physics benchmarks and auditing them with experts, focusing on text-only problems with verifiable final answers. For each subfield of physics, faculty and graduate researchers with relevant expertise carefully review problem statements, reference solutions, and model responses to distinguish genuine model errors from grader errors, incorrect reference solutions, and ambiguous or underspecified questions. Most audited cases initially evaluated as incorrect reflect these benchmarking issues rather than errors in the models' physics reasoning. We then ask experts to address these benchmarking issues by correcting erroneous reference solutions and repairing or excluding flawed questions. We find that GPT-5.6-Sol's measured mean@4 rises from 47.3% to 78.7% on HLE-Physics and from 61.0% to 87.2% on CMT-Benchmark, while its corrected pass@4 reaches 94.4% on the 54 retained CritPt challenges. Corrected scores are computed on the retained evaluation subsets following expert review. Scores on the audited subsets of UGPhysics, PRISM-Physics, and PHYBench also rise substantially after correction. These findings suggest that current benchmarks substantially understate frontier models' ability to solve well-posed physics problems. Near-saturation on these closed-ended tasks highlights the need for more demanding, expert-validated evaluations.

cs.AI↗

arXiv2026

Testing General Relativity with Individual Supermassive Black Hole Binaries

We develop a unified framework for testing gravity beyond General Relativity (GR) with continuous gravitational waves (CWs) from individual supermassive black hole binaries (SMBHBs). These long-lived, nearly monochromatic nanohertz signals offer unique strengths for precision tests of gravity, since their coherent phase evolution and inter-pulsar correlations in pulsar timing arrays (PTAs) retain detailed information about departures from GR over cosmological propagation distances. We consider three representative classes of deviations from GR: additional polarization states, modified dispersion relations, and parity-violating birefringence. For each, we derive the inter-pulsar cross correlation, the modified antenna response, and the propagation-induced pulsar-term phase delay. For non-tensorial polarizations, the CW cross correlation scales linearly in the alternative-polarization amplitude, compared to the quadratic scaling of the gravitational-wave background (GWB), provided the beyond-GR modes are sub-dominant. PTAs are also competitive for modified dispersion relations, where low frequencies enhance both the antenna-pattern modification and the pulsar-term phase delay. Birefringence, by contrast, is suppressed at nanohertz frequencies for most parity-violating theories. We validate the framework with injection-and-recovery simulations for breathing-mode and massive-graviton signals at current observational limits, recovering the injected beyond-GR parameters and distinguishing the CW signal from both correlated and uncorrelated background models. We further show that a pure-GR CW template recovers source parameters without significant bias when beyond-GR physics is present in the data, supporting a two-stage analysis strategy: identify candidates under GR, then test for deviations.

gr-qc↗

arXiv2026

Fingerprints of Individual Supermassive Black Hole Binaries in Pulsar Timing Arrays

With evidence for a nanoHertz gravitational-wave background now established by Pulsar Timing Arrays, the search focuses on identifying individual supermassive black hole binaries. We show that these binaries produce a distinct spatial correlation pattern across the array, acting as a deterministic analogue to the stochastic Hellings \& Downs curve. We derive a closed analytic expression for this single-source overlap reduction function, $Υ_{ab}$, factorizing the signal into a source-dependent amplitude and a purely geometric fingerprint. Using simulated datasets, we demonstrate that this fingerprint breaks the degeneracy between an individual binary and a stochastic background. Including these cross-correlations yields Bayes factors of $1611$ favoring the continuous-wave model over a Hellings \& Downs correlated background model and $159$ favoring the continuous-wave model over an uncorrelated red-noise model. Furthermore, these new cross-correlations improve sky localization by a factor of $11\times$ over an uncorrelated search. Finally, while coherent matched filtering offers higher theoretical sensitivity, we argue that a cross-correlation-based search for individual binaries provides a robust alternative that hedges against the possibility of overfitting to noise fluctuations by focusing on the evidence for the correlations. Indeed, the geometric fingerprints we present here show that spatial correlations can also be used to identify the first nanoHertz gravitational-wave sources.

astro-ph.HE↗

arXiv2026

Graviton Photoproduction by a Kerr-Newman Black Hole with Worldline EFT

We present the first computation of the gauge-invariant, long-wavelength scattering amplitude for the graviton photoproduction by a Kerr-Newman black hole through $\mathcal{O}\big((ωm)^2\big)$, or correspondingly $\mathcal{O}(S^2)$, and to linear order in $G$, using the worldline effective field theory. We show that electromagnetic interactions can be introduced consistently into the spinning worldline theory while preserving spin gauge invariance. We also derive the full angular dependence of the conversion cross section through $\mathcal{O}(S^2)$, and demonstrate that the relevant Wilson coefficients at this order are fixed entirely by matching the electromagnetic and gravitational multipole moments to the Kerr-Newman solution. This result provides a benchmark for future analyses of coupled gravitoelectromagnetic scattering in spinning, charged compact-object backgrounds.

hep-th↗

arXiv2026

QHap: Quantum-Inspired Haplotype Phasing

Haplotype phasing, the process of resolving parental allele inheritance patterns in diploid genomes, is critical for precision medicine and population genetics, yet the underlying optimization is NP-hard, posing a scalability challenge. To address this, we introduce QHap, a haplotype phasing algorithm that leverages quantum-annealing-inspired optimization. By reformulating haplotype phasing as a Max-Cut problem and deploying a GPU-accelerated ballistic simulated bifurcation solver, QHap accelerates phasing while maintaining accuracy comparable to established phasing tools. On the highly polymorphic human major histocompatibility complex region, QHap demonstrates 4- to 20-fold acceleration over HapCUT2 and WhatsHap with zero switch error across multiple long-read sequencing platforms. The framework implements two strategies: a read-based method for regional phasing, and a single nucleotide polymorphism-based method that, through quality-weighted probabilistic edge construction, efficiently scales to chromosome-scale tasks. Integration of Pore-C chromatin conformation capture data increases the haplotype N50 by up to 15-fold, enabling near-chromosome-scale haplotype reconstruction. QHap demonstrates that quantum-inspired algorithms operating on classical hardware offer a promising approach to addressing the growing computational demands of sequencing data, establishing a new paradigm for applying physics-inspired optimization to fundamental challenges in computational genomics.

q-bio.GN↗

arXiv2026

Multi-Static ISAC Assisted by Double-Side Fluid Antenna System

As a pivotal usage scenario for 6G networks, integrated sensing and communication (ISAC) has emerged as a focal point of both academic and industrial research. To accommodate the heterogeneous connectivity requirements of future networks while jointly enhancing both the sensing and communication performance, this paper integrates the multi-static ISAC architecture with double-side fluid antenna system (DS-FAS) to fully exploit the available spatial degrees-of-freedom. Specifically, we establish a joint optimization framework for FA positions and transmit beamforming to maximize the target detection probability while satisfying the communication quality-of-service requirements. Recognizing the intricate coupling between the double-side FA positions and transmit beamforming, instead of trying to obtain an initial feasible point, we resort to the penalty-based mechanism to ensure the robustness against initial feasibility without introducing additional non-convexity. An alternating optimization-based algorithm is proposed to solve the decoupled subproblems. Specifically, the transmit beamforming is globally optimized via the semidefinite relaxation technique, while the transmit FA positions are determined using the majorization-minimization method. Finally, leveraging the analyzed FA mechanism, the feasibility subproblem for receive FA positions is transformed into a signal-to-interference-plus-noise ratio maximization one, solved efficiently via a gradient ascent-based approach, which yields superior performance over the feasibility-based benchmark with reduced complexity. Numerical results demonstrate the superiority of the considered DS-FAS-assisted multi-static ISAC systems in both noise-limited and interference-limited scenarios, while key insights for practical deployment are further extracted from the simulation analysis.

eess.SP↗

arXiv2026

Gravity Echoes from Supermassive Black Hole Binaries

Pulsar timing arrays record gravitational waves from supermassive black hole binaries at two spacetime points: an Earth term, measured when the wave passes the Earth, and a pulsar term, measured when the wave passed each pulsar at an earlier epoch. We show that a future $μ$Hz-band detection of a nearby massive binary by a mission such as $μ$Ares would turn PTA pulsar terms into targeted probes of binary evolution. In analogy with supernova light echoes, each pulsar term acts as a gravity echo: a dated snapshot of the binary at an earlier stage of its inspiral. Together, the $μ$Hz Earth-term measurement and the nHz pulsar-term echoes provide a temporal baseline that neither detector could access alone. For a fiducial equal-mass binary with total mass $10^9\,M_\odot$ at 80~Mpc, we find a combined pulsar timing array echo signal-to-noise ratio of 33, with up to 24 pulsars individually resolving the signal among pulsars with 50-year baselines. The angular dependence of the single-pulsar echo sensitivity alone enables independent sky localization of the source to $\sim$10--100~deg$^2$, and the resolved pulsar-term frequencies directly measure the binary inspiral rate hundreds to thousands of years ago. With sufficient pulsar distance precision, a small set of anchor pulsars could additionally phase-connect the array and trace the post-Newtonian evolution coherently over kpc baselines. The source population required for gravity echoes is drawn from the same massive-end census responsible for the observed nanoHertz stochastic background.

astro-ph.HE↗

arXiv2026

Enhanced Distributed Variational Quantum Eigensolver for Large-Scale MaxCut Problem

MaxCut is a canonical NP-hard combinatorial optimization problem in graph theory with broad applications ranging from physics to bioinformatics. Although variational quantum algorithms offer promising new approaches that may eventually outperform classical schemes, they suffer from resource constraints and trainability issues such as barren plateaus, making large-scale instances intractable on noisy intermediate-scale quantum devices. In this paper, we propose an enhanced distributed variational quantum eigensolver for large-scale MaxCut problems, which extends our prior distributed variational quantum eigensolver framework by integrating a novel hybrid classical-quantum perturbation strategy, enhances optimization scalability and efficiency. Our algorithm solves weighted MaxCut instances with up to 1000 vertices using only 10 qubits, and numerical results indicate that it consistently outperforms the Goemans-Williamson algorithm. We further employ a warm-start initialization strategy, seeding the algorithm with high-quality solutions from the Goemans-Williamson algorithm, with results confirming that the optimal classical solution can be effectively further improved. The practical utility of the proposed algorithm is further validated through its application to haplotype phasing on genome sequencing data of the human ABCA1 gene, producing high-quality haplotypes that rival those obtained by the Goemans-Williamson algorithm with $10^6$ projections. These results establish the proposed algorithm as a scalable, NISQ-compatible framework for near-term quantum-enhanced large-scale combinatorial optimization.

quant-ph↗

arXiv2025

Probing Picohertz Gravitational Waves with Pulsars

With periods much longer than the duration of current pulsar timing surveys, gravitational waves in the picohertz (pHz) regime are not detectable in the typical analysis framework for pulsar timing data. However, signatures of these low-frequency signals persist in the slow variation of pulsar timing parameters. In this work, we present the results of the first Bayesian search for continuous pHz gravitational waves using the drift of two sensitive pulsar timing parameters -- time derivative of pulsar binary orbital period $\dot{P}_b$ and second order time derivative of pulsar spin period $\ddot{P}$. We apply our new technique to a dataset with more than double the number of pulsars as previous searches in this frequency band, achieving an order-of-magnitude sensitivity improvement. No continuous wave signal is detected in current data; however, we show that future observations by the Square Kilometre Array will provide significantly improved sensitivity and the opportunity to observe continuous pHz signals, including the early stages of supermassive black hole mergers. We explore the detection prospects for this signal by extending existing population models into the pHz regime, finding that future observations will probe phenomenologically-interesting parameter space. Our new Bayesian technique and leading sensitivity in this frequency domain paves the way for new discoveries in both black hole astrophysics and the search for new physics in the early universe.

astro-ph.HE↗

arXiv2025

WebGuard: Building a Generalizable Guardrail for Web Agents

The rapid development of autonomous web agents powered by Large Language Models (LLMs), while greatly elevating efficiency, exposes the frontier risk of taking unintended or harmful actions. This situation underscores an urgent need for effective safety measures, akin to access controls for human users. To address this critical challenge, we introduce WebGuard, the first comprehensive dataset designed to support the assessment of web agent action risks and facilitate the development of guardrails for real-world online environments. In doing so, WebGuard specifically focuses on predicting the outcome of state-changing actions and contains 4,939 human-annotated actions from 193 websites across 22 diverse domains, including often-overlooked long-tail websites. These actions are categorized using a novel three-tier risk schema: SAFE, LOW, and HIGH. The dataset includes designated training and test splits to support evaluation under diverse generalization settings. Our initial evaluations reveal a concerning deficiency: even frontier LLMs achieve less than 60% accuracy in predicting action outcomes and less than 60% recall in lagging HIGH-risk actions, highlighting the risks of deploying current-generation agents without dedicated safeguards. We therefore investigate fine-tuning specialized guardrail models using WebGuard. We conduct comprehensive evaluations across multiple generalization settings and find that a fine-tuned Qwen2.5VL-7B model yields a substantial improvement in performance, boosting accuracy from 37% to 80% and HIGH-risk action recall from 20% to 76%. Despite these improvements, the performance still falls short of the reliability required for high-stakes deployment, where guardrails must approach near-perfect accuracy and recall.

cs.AI↗

arXiv2025

Unit-density SU(3) Fermi-Hubbard Model with Spin Flavor Imbalance

The advent of ultracold alkaline-earth atoms in optical lattices has established a platform for investigating correlated quantum matter with SU($N$) symmetry, offering highly tunable model parameters that allow experiments to access phenomena that are unavailable in conventional materials. Understanding the ground-state physics of SU($N$) Fermi-Hubbard models away from the Heisenberg limit and from the spin-flavor balanced setting is important, as examining the flavor imbalance reveals new physics in Fermi-Hubbard models and shows how SU($N$) phases react to practical experimental imperfections in optical lattices. In this study, mean-field phase diagrams are presented for the unit-density SU(3) Fermi-Hubbard model at two sets of flavor densities, $\left(\tfrac{1}{3}-δ,\tfrac{1}{3}+δ,\tfrac{1}{3}\right)$ and $\left(\tfrac{1}{4}-δ,\tfrac{1}{4}+δ,\tfrac{1}{2}\right)$, with the flavor imbalance introduced as $δ$. Novel phases are identified at moderate interaction strengths for both densities and their robustness is investigated in the presence of flavor imbalance. Furthermore, we provide microscopic explanations of the phases found and their stability. Analysis of thermal ensembles of random mean-field solutions indicate that, at temperatures accessible in state-of-the-art cold atom experiments, some spin orders are hard for conventional scattering or local observable measurements to detect, but can be more accessible with quantum gas microscopy in optical lattice experiments. This work also shows that nesting and Mottness, intertwined in the usual SU(2) Hubbard model in stark contrast to generic materials, can be tuned in the SU(3) model and play distinct roles. The resulting phase diagrams not only deepen our understanding of SU($N$) Fermi-Hubbard models but also inform future experimental search for new phases.

cond-mat.quant-gas↗