Hui Dai, Zachary Geary, Leo P. Kadanoff
A Toeplitz matrix is one in which the matrix elements are constant along diagonals. The Fisher-Hartwig matrices are much-studied singular matrices in the Toeplitz family. The matrices are defined for all orders, $N$. They are parametrized by two constants, $α$ and $β$. Their spectrum of eigenvalues has a simple asymptotic form in the limit as $N$ goes to infinity. Here we study the structure of their eigenvalues and eigenvectors in this limiting case. We specialize to the case $0<α<|β|<1$, where the behavior is particularly simple.
Yu Xu, Jin Lin, Yu-Huai Li, Hui Dai, Sheng-Kai Liao, Cheng-Zhi Peng
A variable-delay optical interferometer with 128 actively selectable delays and a technique of active phase stabilization are innovatively designed and applied for the first time in the experiment of round-robin differential phase shift quantum key distribution (RRDPS-QKD). According to the RRDPS protocol, larger number of delay channels in interferometer can ensure higher tolerance of bit errors, eliminating the fundamental threshold of bit error rate of 11% of traditional BB84 protocol. Considering that the implementation of RRDPS protocol relys on the realizaiton of the interferometer, therefore, an interferometer with 128 selectable delay paths is constructed and demands the ability of fast switching at the rate of 10 kHz, which requires dynamic stability of multiple paths. Thus, a specific designed phase stabilization technique with closed real time feedback loop is introduced to guarantee the high visibility of interferometer selections dynamically. The active phase stabilization technique employs a phase modulator (PM) driven by a DAC to adjust the relative phase between the two arms of the interferometer. By monitoring photon counting rates of two Up-Conversion Detectors (UCD) at two output ports of the interferometer, a Field Programmable Gate Array (FPGA) calculates and finds the optimal code value for the DAC, maintaining a high visibility of the interferometer every time a new light path is selected. The visibility of most of the 128 interferometer selections can simultaneously be maintained over 96% during the QKD, which lays the foundation for the RRDPS experiment.
Juan Yin, Yuan Cao, Yu-Huai Li, Sheng-Kai Liao, Liang Zhang, Ji-Gang Ren, Wen-Qi Cai, Wei-Yue Liu, Bo Li, Hui Dai, Guang-Bing Li, Qi-Ming Lu, Yun-Hong Gong, Yu Xu, Shuang-Lin Li, Feng-Zhi Li, Ya-Yun Yin, Zi-Qing Jiang, Ming Li, Jian-Jun Jia, Ge Ren, Dong He, Yi-Lin Zhou, Xiao-Xiang Zhang, Na Wang, Xiang Chang, Zhen-Cai Zhu, Nai-Le Liu, Yu-Ao Chen, Chao-Yang Lu, Rong Shu, Cheng-Zhi Peng, Jian-Yu Wang, Jian-Wei Pan
Long-distance entanglement distribution is essential both for foundational tests of quantum physics and scalable quantum networks. Owing to channel loss, however, the previously achieved distance was limited to ~100 km. Here, we demonstrate satellite-based distribution of entangled photon pairs to two locations separated by 1203 km on the Earth, through satellite-to-ground two-downlink with a sum of length varies from 1600 km to 2400 km. We observe a survival of two-photon entanglement and a violation of Bell inequality by 2.37+/-0.09 under strict Einstein locality conditions. The obtained effective link efficiency at 1200 km in this work is over 12 orders of magnitude higher than the direct bidirectional transmission of the two photons through the best commercial telecommunication fibers with a loss of 0.16 dB/km.
Hui Dai, Dan Pechi, Xinyi Yang, Garvit Banga, Raghav Mantri
The Needle-in-a-haystack (NIAH) test is a general task used to assess language models' (LMs') abilities to recall particular information from long input context. This framework however does not provide a means of analyzing what factors, beyond context length, contribute to LMs' abilities or inabilities to separate and recall needles from their haystacks. To provide a systematic means of assessing what features contribute to LMs' NIAH capabilities, we developed a synthetic benchmark called DENIAHL (Data-oriented Evaluation of NIAH for LLM's). Our work expands on previous NIAH studies by ablating NIAH features beyond typical context length including data type, size, and patterns. We find stark differences between GPT-3.5 and LLaMA 2-7B's performance on DENIAHL, and drops in recall performance when features like item size are increased, and to some degree when data type is changed from numbers to letters. This has implications for increasingly large context models, demonstrating factors beyond item-number impact NIAH capabilities.
Hui Dai, Ryan Teehan, Mengye Ren
Many existing evaluation benchmarks for Large Language Models (LLMs) quickly become outdated due to the emergence of new models and training data. These benchmarks also fall short in assessing how LLM performance changes over time, as they consist of a static set of questions without a temporal dimension. To address these limitations, we propose using future event prediction as a continuous evaluation method to assess LLMs' temporal generalization and forecasting abilities. Our benchmark, Daily Oracle, automatically generates question-answer (QA) pairs from daily news, challenging LLMs to predict "future" event outcomes. Our findings reveal that as pre-training data becomes outdated, LLM performance degrades over time. While Retrieval Augmented Generation (RAG) has the potential to enhance prediction accuracy, the performance degradation pattern persists, highlighting the need for continuous model updates. Code and data are available at https://agenticlearning.ai/daily-oracle.
Sheng-Kai Liao, Hai-Lin Yong, Chang Liu, Guo-Liang Shentu, Dong-Dong Li, Jin Lin, Hui Dai, Shuang-Qiang Zhao, Bo Li, Jian-Yu Guan, Wei Chen, Yun-Hong Gong, Yang Li, Ze-Hong Lin, Ge-Sheng Pan, Jason S. Pelc, M. M. Fejer, Wen-Zhuo Zhang, Wei-Yue Liu, Juan Yin, Ji-Gang Ren, Xiang-Bin Wang, Qiang Zhang, Cheng-Zhi Peng, Jian-Wei Pan
Nov 30, 2016·quant-ph·PDF Satellite based quantum communication has been proven as a feasible way to achieve global scale quantum communication network. Very recently, a low-Earth-orbit (LEO) satellite has been launched for this purpose. However, with a single satellite, it takes an inefficient 3-day period to provide the worldwide connectivity. On the other hand, similar to how the Iridium system functions in classic communication, satellite constellation (SC) composed of many quantum satellites, could provide global real-time quantum communication. In such a SC, most of the satellites will work in sunlight. Unfortunately, none of previous ground testing experiments could be implemented at daytime. During daytime, the bright sunlight background prohibits quantum communication in transmission over long distances. In this letter, by choosing a working wavelength of 1550 nm and developing free-space single-mode fibre coupling technology and ultralow noise up-conversion single photon detectors, we overcome the noise due to sunlight and demonstrate a 53-km free space quantum key distribution (QKD) in the daytime through a 48-dB loss channel. Our system not only shows the feasibility of satellite based quantum communication in daylight, but also has the ability to naturally adapt to ground fibre optics, representing an essential step towards a SC-based global quantum network.
Rongge Xu, Hui Dai, Yiming Fu, Jiedong Jiang, Tianjiao Nie, Junkai Wang, Holiverse Yang, Zhi-Hao Zhang
While large language models (LLMs) have demonstrated impressive capabilities in formal theorem proving, current benchmarks fail to adequately measure library-grounded abstraction -- the ability to reason with high-level interfaces and reusable structures central to modern mathematics and software engineering. We introduce LeanCat, a challenging benchmark comprising 100 fully formalized category-theory tasks in Lean. Unlike algebra or arithmetic, category theory serves as a rigorous stress test for structural, interface-level reasoning. Our evaluation reveals a severe abstraction gap: the best state-of-the-art model solves only 12.0% of tasks at pass@4, with performance collapsing from 55.0% on Easy tasks to 0.0% on High-difficulty tasks, highlighting a failure in compositional generalization. To overcome this, we evaluate LeanBridge, a retrieval-augmented agent that employs a retrieve-generate-verify loop. LeanBridge achieves a peak success rate of 24.0% -- doubling the performance of the best static baseline. These results empirically demonstrate that iterative refinement and dynamic library retrieval are not merely optimizations but strict necessities for neuro-symbolic reasoning in abstract domains. LeanCat offers a compact, reusable testbed for tracking progress toward reliable, research-level formalization.
Hui Dai, Qi Shen, Chao-Ze Wang, Shuang-Lin Li, Wei-Yue Liu, Wen-Qi Cai, Sheng-Kai Liao, Ji-Gang Ren, Juan Yin, Yu-Ao Chen, Qiang Zhang, Feihu Xu, Cheng-Zhi Peng, Jian-Wei Pan
High-precision time synchronization for remote clocks plays an important role in fundamental science and real-life applications. However, the current time synchronization techniques have been shown to be vulnerable to sophisticated adversaries. There is a compelling need for fundamentally new methods to distribute high-precision time information securely. Here we propose a satellite-based quantum-secure time transfer (QSTT) scheme based on two-way quantum key distribution (QKD) in free-space, and experimentally verify the key technologies of the scheme via the Micius quantum satellite. In QSTT, a quantum signal (e.g., single photon) is used as the carrier for both the time transfer and the secret-key generation, offering quantum-enhanced security for transferring time signal and time information. We perform a satellite-to-ground time synchronization using single-photon-level signals and achieve a quantum bit error rate of less than 1%, a time data rate of 9 kHz and a time-transfer precision of 30 ps. These results offer possibilities towards an enhanced infrastructure of time-transfer network, whose security stems from quantum physics.
Yu-Huai Li, Yuan Cao, Hui Dai, Jin Lin, Zhen Zhang, Wei Chen, Yu Xu, Jian-Yu Guan, Sheng-Kai Liao, Juan Yin, Qiang Zhang, Xiongfeng Ma, Cheng-Zhi Peng, Jian-Wei Pan
May 29, 2015·quant-ph·PDF In conventional quantum key distribution (QKD) protocols, security is guaranteed by estimating the amount of leaked information through monitoring signal disturbance, which, in practice, is generally caused by environmental noise and device imperfections rather than eavesdropping. Such estimation therefore tends to overrate the amount of leaked information in practice, leads to a fundamental threshold of the bit error rate. The threshold becomes a bottleneck of the development of practical QKD systems. In classical communication, according to Shannon's communication theory, information can transform through a noisy channel even if the background noise is very strong compare to the signal and hence the threshold of the bit error rate tends to 50%. One might wonder whether a QKD scheme can also tolerate error rate as high as 50%. The question is answered affirmatively with the recent work of round-robin differential phase-shift (RRDPS) protocol, which breaks through the fundamental threshold of the bit error rate and indicates another potential direction in the field of quantum cryptography. The key challenge to realize the RRDPS scheme lies on the measurement device, which requires a variable-delay interferometer. The delay needs to be chosen from a set of predetermined values randomly. Such measurement can be realized by switching between many interferometers with different delays at a high speed in accordance with the system repetition rate. The more delay values can be chosen from, the higher error rate can be tolerated. By designing an optical system with multiple switches and employing an active phase stabilization technology, we successfully construct a variable-delay interferometer with 128 actively selectable delays. With this measurement, we experimentally demonstrate the RRDPS QKD protocol and obtain a final key rate of 15.54 bps via a total loss of 18 dB and 8.9% error rate.
James Mullenbach, Yada Pruksachatkun, Sean Adler, Jennifer Seale, Jordan Swartz, T. Greg McKelvey, Hui Dai, Yi Yang, David Sontag
Continuity of care is crucial to ensuring positive health outcomes for patients discharged from an inpatient hospital setting, and improved information sharing can help. To share information, caregivers write discharge notes containing action items to share with patients and their future caregivers, but these action items are easily lost due to the lengthiness of the documents. In this work, we describe our creation of a dataset of clinical action items annotated over MIMIC-III, the largest publicly available dataset of real clinical notes. This dataset, which we call CLIP, is annotated by physicians and covers 718 documents representing 100K sentences. We describe the task of extracting the action items from these documents as multi-aspect extractive summarization, with each aspect representing a type of action to be taken. We evaluate several machine learning models on this task, and show that the best models exploit in-domain language model pre-training on 59K unannotated documents, and incorporate context from neighboring sentences. We also propose an approach to pre-training data selection that allows us to explore the trade-off between size and domain-specificity of pre-training datasets for this task.
Chaohong Lee, Jiahao Huang, Haiming Deng, Hui Dai, Jun Xu
In quantum interferometry, it is vital to control and utilize nonlinear interactions for achieving high-precision measurements. Attribute to their long coherent time and high controllability, ultracold atoms including Bose condensed atoms have been widely used for implementing quantum interferometry. Here, we review the recent progresses in theoretical studies of quantum interferometry with Bose condensed atoms. In particular, we focus on the nonlinear phenomena induced by the atom-atom interaction and how to control and utilize these nonlinear phenomena. Under the mean-field description, due to the atom-atom interaction, matter-wave solitons appear in the interference patterns, and macroscopic quantum self-trapping exists in the Bose-Josephson junctions. Under the many-body description, the atom-atom interaction can generate non-classical entanglement, which may be utilized to achieve high-precision measurements beyond the standard quantum limit.
Hui Dai, Dirk K. Morr
We study the non-equilibrium transport properties of a one-dimensional array of dissipative quantum dots. Using the Keldysh formalism, we show that the dots' dissipative nature leads to a spatial variation of the chemical potential, which in disordered arrays, breaks the invariance of the current, I, under bias reversal. Moreover, the array's nanoscopic size results in an algebraic low-temperature dependence of I. Finally, we show that a local Coulomb interaction splits the dots' electronic levels, resulting in a Coulomb blockade, which is softened with increasing dissipation and array size.
James Mullenbach, Jordan Swartz, T. Greg McKelvey, Hui Dai, David Sontag
Both electronic health records and personal health records are typically organized by data type, with medical problems, medications, procedures, and laboratory results chronologically sorted in separate areas of the chart. As a result, it can be difficult to find all of the relevant information for answering a clinical question about a given medical problem. A promising alternative is to instead organize by problems, with related medications, procedures, and other pertinent information all grouped together. A recent effort by Buchanan (2017) manually defined, through expert consensus, 11 medical problems and the relevant labs and medications for each. We show how to use machine learning on electronic health records to instead automatically construct these problem-based groupings of relevant medications, procedures, and laboratory tests. We formulate the learning task as one of knowledge base completion, and annotate a dataset that expands the set of problems from 11 to 32. We develop a model architecture that exploits both pre-trained concept embeddings and usage data relating the concepts contained in a longitudinal dataset from a large health system. We evaluate our algorithms' ability to suggest relevant medications, procedures, and lab tests, and find that the approach provides feasible suggestions even for problems that are hidden during training. The dataset, along with code to reproduce our results, is available at https://github.com/asappresearch/kbc-pomr.
Seung-Yeop Lee, Hui Dai, Eldad Bettelheim
We study the asymptotic eigenvalue distribution of Toeplitz matrices generated by a singular symbol. It has been conjectured by Widom that, for a generic symbol, the eigenvalues converge to the image of the symbol. In this paper we ask how the eigenvalues converge to the image. For a given Toeplitz matrix $T_n(a)$ of size $n$, we take the standard approach of looking at $\det(ζ-T_n(a))$, of which the asymptotic information is given by the Fisher-Hartwig theorem. For a symbol with single jump, we obtain the distribution of eigenvalues as an expansion involving $1/n$ and $\log n/n$. To demonstrate the validity of our result we compare our result against the numerics using a pure Fisher-Hartwig symbol.
Tankut Can, Hui Dai, Dirk K. Morr
Using the non-equilibrium Keldysh Green's function formalism, we show that the non-equilibrium charge transport in nanoscopic quantum networks takes place via {\it current eigenmodes} that possess characteristic spatial patterns. We identify the microscopic relation between the current patterns and the network's electronic structure and topology and demonstrate that these patterns can be selected via gating or constrictions, providing new venues for manipulating charge transport at the nanoscale. Finally, decreasing the dephasing time leads to a smooth evolution of the current patterns from those of a ballistic quantum network to those of a classical resistor network.
Yang You, Laming Chen, Yuantao Gu, Wei Feng, Hui Dai
A new sparse signal recovery algorithm for multiple-measurement vectors (MMV) problem is proposed in this paper. The sparse representation is iteratively drawn based on the idea of zero-point attracting projection (ZAP). In each iteration, the solution is first updated along the negative gradient direction of an approximate $\ell_{2,0}$ norm to encourage sparsity, and then projected to the solution space to satisfy the under-determined equation. A variable step size scheme is adopted further to accelerate the convergence as well as to improve the recovery accuracy. Numerical simulations demonstrate that the performance of the proposed algorithm exceeds the references in various aspects, as well as when applied to the Modulated Wideband Converter, where recovering MMV problem is crucial to its performance.
Lei Wang, Hang Liu, Junwei Liu, Aoxuan Liu, Jialiang Huang, Qiannan Li, Hui Dai, Caihong Zhang, Jingbo Wu, Kebin Fan, Huabing Wang, Biaobing Jin, Jian Chen, Peiheng Wu
Bound states in the continuum (BICs) and exceptional points (EPs), as two distinct physical singularities represented by complex frequencies in non-Hermitian systems, have garnered significant attention and clear definitions in their respective fields in recent years. They share overlapping applications in areas such as high-sensitivity sensing and laser emission. However, the transition between the two, inspired by these intersections, remains largely unexplored. In this work, we reveal the transition process in a non-Hermitian two-mode system, evolving from one bound singularity to a two-dimensional exceptional ring, where the EP is the coalescent state of the quasi-Friedrich-Wintgen (FW)-BIC. This phenomenon is experimentally validated through pored dielectric metasurfaces in terahertz band. Furthermore, external pumping induced photocarriers as the dissipative perturbation, facilitates the breaking of degeneracy in the complex eigenfrequency and enables dynamic EP switching. Finally, we experimentally demonstrate a switchable terahertz beam deflection driven by the phase singularities of the EP. These findings are instrumental in advancing the development of compact devices for sensing and wavefront control within non-Hermitian systems.
Tao Lei, Yu Zhang, Sida I. Wang, Hui Dai, Yoav Artzi
Common recurrent neural architectures scale poorly due to the intrinsic difficulty in parallelizing their state computations. In this work, we propose the Simple Recurrent Unit (SRU), a light recurrent unit that balances model capacity and scalability. SRU is designed to provide expressive recurrence, enable highly parallelized implementation, and comes with careful initialization to facilitate training of deep models. We demonstrate the effectiveness of SRU on multiple NLP tasks. SRU achieves 5--9x speed-up over cuDNN-optimized LSTM on classification and question answering datasets, and delivers stronger results than LSTM and convolutional models. We also obtain an average of 0.7 BLEU improvement over the Transformer model on translation by incorporating SRU into the architecture.
Haiming Deng, Hui Dai, Jiahao Huang, Xizhou Qin, Jun Xu, Honghua Zhong, Chunshan He, Chaohong Lee
We present a cluster Gutzwiller mean-field study for ground states and time-evolution dynamics in the Bose-Hubbard ladder (BHL), which can be realized by loading Bose atoms in double-well optical lattices. In our cluster mean-field approach, we treat each double-well unit of two lattice sites as a coherent whole for composing the cluster Gutzwiller ansatz, which may remain some residual correlations in each two-site unit. For a unbiased BHL, in addition to conventional superfluid phase and integer Mott insulator phases, we find that there are exotic fractional insulator phases if the inter-chain tunneling is much stronger than the intra-chain one. The fractional insulator phases can not be found by using a conventional mean-field treatment based upon the single-site Gutzwiller ansatz. For a biased BHL, we find there appear single-atom tunneling and interaction blockade if the system is dominated by the interplay between the on-site interaction and the inter-chain bias. In the many-body Landau-Zener process, in which the inter-chain bias is linearly swept from negative to positive or vice versa, our numerical results are qualitatively consistent with the experimental observation [Nat. Phys. \textbf{7}, 61 (2011)]. Our cluster bosonic Gutzwiller treatment is of promising perspectives in exploring exotic quantum phases and time-evolution dynamics of bosonic particles in superlattices.