arXiv Science⌕ Search

arXiv subjects

Xueting Fang

Publications and source records attributed to Xueting Fang.

3 recordsLinked to original sources

Competition and cancellation in a tilted Rice--Mele chain: exact Wannier--Stark offset and spectral readout

In a tilted Rice--Mele chain, the Wannier--Stark (WS) ladder becomes staggered, \(E_n=n\Fo+(-1)^{n+1}A\). The offset \(A\) reflects the competition between the staggered onsite energy \(Δ\) and the dimerized hopping \(\dt\), and its cancellation restores spectral equidistance. We determine \(A\) through a systematic strong-field series in the WS basis and a nonperturbative two-band Floquet expression. An exact matrix element reduces the strong-field expansion to Bessel algebra; at the representative working point, truncation at seventh order locates the cancellation point with a theory-internal truncation error of \(0.010\)~Hz, against \(27.3\)~Hz at first order. The Floquet propagator gives the closed form \(A=(\Fo/π)\arcsin(-i\,\mathrm{Tr}\,U/2)\), exact within the two-band model, implying the kinematic bound \(|A|\le\Fo/2\) and identifying cancellation with maximum spectral uniformity. The same analytic structure yields a family of real zeros and complex exceptional points that set the convergence radius of the strong-field series without touching the real axis. On odd-length chains, the hidden-symmetry defect produces a pairing plateau of height \(2|A|\) that persists numerically beyond first order in the modulations, allowing the offset to be read from the static spectrum without Floquet propagation or time evolution. In the bulk, cancellation appears as the merging of two sidebands into a single line at \(\Fo\), a null measurement that requires no calibrated sideband amplitudes, although the tilt and line frequency still need calibration. These signatures provide complementary spectral tests of the staggered WS ladder and a calibration-aware route to precision force sensing.

quant-ph↗

KlinikeBench: Evaluating Language Models Beyond Diagnostic Accuracy

Most clinical benchmarks evaluate language models (LMs) on diagnosis using complete case descriptions. In clinical practice, however, patients present information in different ways, and clinicians must obtain relevant history and determine which examinations are needed before reaching a diagnosis. Diagnostic accuracy alone therefore cannot establish whether an agent gathered essential information or conducted an appropriate clinical assessment. Furthermore, existing benchmarks lack professional clinicians' verification. To address this gap, we introduce KlinikeBench, a benchmark of 333 clinician-authored tasks, each providing an isolated sandbox environment with a virtual patient, clinical tools, and task-specific success criteria. More than 35 clinicians contributed to case authoring and benchmark evaluation. In an empirical study, clinicians gave simulated dialogues higher mean quality ratings than reference conversations, which is adapted from real conversation. In each task, an LM has a fixed budget of turns to communicate with the patient, ask about relevant history, request examinations, follow action constraints, and record a final diagnosis. We score these steps separately as well as together. Across 31 models and seven model families, the best-performing models (e.g., GPT-6-astra and Claude Opus 5) succeed on less than 30% of tasks, even though their diagnosis accuracy reaches 90.7%. Some models benefit from talking with the patient; others diagnose well from a complete chart but perform much worse in conversation. Overall, KlinikeBench provides a testbed for evaluating the full clinical encounter and reveals a substantial gap between diagnostic accuracy and performance in interactive clinical assessment. All the code and data is available on https://zehui127.github.io/klinikebench/

cs.CL↗

Optimized Slice-Phase Control of Mirror Pulse in Cold-Atom Interferometry with Finite Response Time

Atom interferometers require both high efficiency and robust performance in their mirror pulses under experimental inhomogeneities. In this work, we demonstrated that quantum optimal control designed mirror pulse significantly enhance interferometer performance by using novel adaptive sliced structure. Using gradient ascent pulse engineering (GRAPE), optimized mirror pulse for a Mach-Zehnder light-pulse atom interferometer was designed by discretizing the control into non-uniform phase slices. This design broadened the tolerence to experimentally relevant variations in detuning $[-Ω_0,Ω_0]$ and Rabi frequency $[0.1\timesΩ_0,1.9\timesΩ_0]$ ($Ω_0=2π\times25$ kHz), while maintaining high transfer efficiency even when the response-time delays up to 1.6 $\rm{μs}$. The optimized pulse was found to be robust to coupling inhomogeneity and velocity spread, offering a significant improvement in robustness over conventional pulse. The adaptive pulse slicing method provides a minimalist strategy that reduces experimental complexity while enhancing robustness and scalability, offering an innovative scheme for quantum optimal control in high precision atom interferometry.

quant-ph↗