arXiv subjects

Chao Xiang

Publications and source records attributed to Chao Xiang.

At least 19 recordsLinked to original sources

arXiv2026

Autonomous discovery of traffic laws with AI traffic scientists

Universal traffic laws describe recurrent patterns in congestion, mobility and driving behavior across cities, providing a scientific basis for transportation planning, management and control. Their discovery, however, remains expert-driven, requiring candidate regularities to be identified from heterogeneous observational evidence or validated through intervention experiments. Although autonomous artificial intelligence (AI) systems have advanced scientific discovery in controlled laboratory settings, extending them to complex transportation domains remains a challenge. Here we present TrafficSci, an agentic AI system that formulates traffic-law discovery as an iterative, auditable workflow integrating evidence scoping, critic-judge hypothesis induction, and observational-interventional validation. Across four case studies spanning population, network, control and trajectory scales, TrafficSci autonomously rediscovers three established traffic laws and identifies an unreported intrinsic temporal memory scale in urban driving behavior, statistically consistent across eight cities and two trajectory datasets. TrafficSci provides a route for extending AI-driven scientific discovery from controlled domains to complex urban systems.

cs.AI↗

arXiv2026

Overcoming noise-agility trade-off in integrated lasers for precision sensing

Lasers that combine narrow linewidths with rapid tunability are critical for applications such as coherent optical ranging, distributed fiber-optic sensing, and precision spectroscopy. Despite significant progress in integrated laser technologies, the concurrent realization of low phase noise and frequency agility on a single integrated platform remains challenging owing to a fundamental architectural trade-off: conventional integrated laser designs typically suppress phase noise via high-$Q$ resonators, yet the extended photon lifetimes inherent to such resonators intrinsically constrain tuning speed. Here, we address this noise-agility trade-off by introducing a laser architecture that achieves ultralow phase noise and ultrafast tunability simultaneously. Rather than relying on ultrahigh-$Q$ resonators for self-injection locking, our design employs strong synthetic feedback within a Pockels-tunable, resonator-enhanced distributed Bragg reflector to suppress phase noise. As a proof of concept, we demonstrate a hybrid integrated laser with a short-term linewidth of 29 Hz, realized using a lithium niobate external cavity with a loaded $Q$ of only 0.62 million. The adoption of a moderate resonator $Q$ relaxes the photon-lifetime constraint on tuning speed, enabling sub-exahertz-per-second tuning rates and a chirp nonlinearity as low as 0.14%. Leveraging this laser, we implement a frequency-modulated continuous-wave LiDAR system that achieves a relative ranging precision of $1.7 \times 10^{-4}$ at a measurement rate of $1\,\text{MSa s}^{-1}$, without requiring complex chirp linearization techniques. We further demonstrate fiber-optic acoustic sensing capable of detecting sub-$\mu\epsilon$ dynamic strain, underscoring the platform's versatility for high-speed precision optical measurements. Our work provides a route toward cost-effective yet high-performance sensing and metrology systems.

physics.optics↗

arXiv2026

Meter-long broadband chirped Bragg gratings for on-chip dispersion control and pulse shaping

Precise on-chip dispersion control is essential for advanced integrated photonic technologies, enabling applications ranging from high-speed communications and sensing to signal processing and biomedical imaging. However, existing on-chip dispersion control methods still suffer from substantial loss and a limited dispersion-bandwidth product (DBP) far from application needs. As a result, on-chip systems continue to rely exclusively on off-chip dispersion control solutions provided by optical fiber or bulky free-space optics. To overcome these limitations, we design and fabricate meter-long chirped spiral Bragg gratings (CSBGs) on the ultra-low-loss silicon nitride (SiN) photonic platform for advanced dispersion control. Our device achieves a 10-nanosecond group delay with customizable bandwidths exceeding 10 nanometers within a compact footprint of only 30 $\text {mm} ^2$, surpassing the physical limits of fiber-based grating devices. More importantly, CSBGs can simultaneously possess the characteristics of high stability, low latency, and a large DBP, thanks to the ultra-low-loss SiN platform with a loss of only 0.3 dB/m. Leveraging the precise and stable dispersion profile, we demonstrate high-fidelity pulse shaping and compression of electro-optic frequency combs (EOCs) with a 1-GHz repetition rate centered across the entire reflection bandwidth. The compressed pulse has an on-chip peak (average) power of 21.6 watts (580 milliwatts). Furthermore, we showcase for the first time the application of on-chip pulse-compressed EOC in wavelength-swept coherent anti-Stokes Raman scattering (CARS) microscopy. Our work provides integrated photonics with a long-sought, scalable, and robust solution for high-performance on-chip dispersion control, empowering a new generation of on-chip functionalities.

physics.optics↗

arXiv2026

Scalable 3D silicon nitride photonic interposer for high-density optical interconnects

Modern computing workloads demand energy-efficient, high-bandwidth interconnects, motivating photonic interposers as an alternative to electrical links. Here we demonstrate a compact 3D silicon nitride (SiN) photonic interposer prototype comprising two routing layers, with the 3D routing scheme optimized by a global optimization algorithm. The 3D interposer realizes a fully connected 12-node optical network that reduces the total number of intralayer crossings from 495 for all-planar routing to merely 150 (69.7% reduction), below the theoretical lower bound of 153 for all-planar interconnects. Comparing the two schemes, our 3D design achieves a 45.8% reduction experimentally in the average loss per waveguide. The proposed 3D routing architecture also features inherent symmetry and is scalable to higher node counts, flexible node placements, additional routing layers, and other operating wavelengths, enabling denser, lower-loss photonic interposers for next-generation scale-up and high-performance computing (HPC) systems.

physics.optics↗

arXiv2026

Building reliable 3D photonic integrated circuits and cavities at the wafer scale

Three-dimensional (3D) photonic integrated circuits (PIC) are emerging as an indispensable scheme for high density and multifunctional photonic systems. However, the wafer-scale scaling of PICs towards a 3D configuration is constrained by two key factors: (i) the trade-off between inter-layer taper efficiency and footprint, and (ii) wafer-scale uniformity of inter-layer transition loss. In this work, we introduce etch-back assisted chemical mechanical polishing (E-CMP) to achieve high wafer-scale uniformity of the spacer layer. Moreover, we break the efficiency-footprint trade-off by demonstrating a novel $\kappa$-engineered taper, achieving a reliability metric that is 75\% higher than the traditional linearly tapered structure. Building on these design and fabrication developments, we enable reliable 3D PICs with typical loss of 0.077 and 0.068 dB/cm on two silicon nitride (SiN) waveguide layers and typical 3D transition loss as low as 6 mdB. Furthermore, the low 3D transition loss enables the first class of 3D high-Q optical cavities occupying two distinct device layers, providing new design space for high-Q optical cavities. The scalable fabrication process and design methodology provide routes for wafer-scale reliable 3D PICs that are promising in a series of applications ranging from photonic interconnects and computing networks to high-density photonic sensors and nonlinear photonics.

physics.optics↗

arXiv2025

BootSeer: Analyzing and Mitigating Initialization Bottlenecks in Large-Scale LLM Training

Large Language Models (LLMs) have become a cornerstone of modern AI, driving breakthroughs in natural language processing and expanding into multimodal jobs involving images, audio, and video. As with most computational software, it is important to distinguish between ordinary runtime performance and startup overhead. Prior research has focused on runtime performance: improving training efficiency and stability. This work focuses instead on the increasingly critical issue of startup overhead in training: the delay before training jobs begin execution. Startup overhead is particularly important in large, industrial-scale LLMs, where failures occur more frequently and multiple teams operate in iterative update-debug cycles. In one of our training clusters, more than 3.5% of GPU time is wasted due to startup overhead alone. In this work, we present the first in-depth characterization of LLM training startup overhead based on real production data. We analyze the components of startup cost, quantify its direct impact, and examine how it scales with job size. These insights motivate the design of Bootseer, a system-level optimization framework that addresses three primary startup bottlenecks: (a) container image loading, (b) runtime dependency installation, and (c) model checkpoint resumption. To mitigate these bottlenecks, Bootseer introduces three techniques: (a) hot block record-and-prefetch, (b) dependency snapshotting, and (c) striped HDFS-FUSE. Bootseer has been deployed in a production environment and evaluated on real LLM training workloads, demonstrating a 50% reduction in startup overhead.

cs.LG↗

arXiv2025

Training Language Model to Critique for Better Refinement

Large language models (LLMs) have demonstrated remarkable evaluation and critique capabilities, providing insightful feedback and identifying flaws in various tasks. However, limited research has explored which types of critiques are most effective for improving model responses or how to generate such critiques. To address this gap, we introduce \textbf{R}efinement-oriented \textbf{C}ritique \textbf{O}ptimization (RCO), a novel framework designed to train critic models using refinement signals. RCO uses a feedback loop where critiques, generated by the critic model, guide the actor model in refining its responses. The critique utility (CU) quantifies the effectiveness of these refinements, serving as the reward signal for training the critic model. By focusing on critiques that lead to better refinements, RCO eliminates the need for direct critique preference assessment, ensuring that critiques driving meaningful improvements are rewarded. We evaluate RCO across five tasks, i.e., dialog generation, summarization, question answering, mathematical reasoning, and code generation, and show that it significantly outperforms traditional methods and open-source models in terms of critique quality and refinement outcomes. Our contributions include the introduction of RCO, a novel supervision scheme based on refined response preferences, and comprehensive experimental results that highlight the method's effectiveness in enhancing LLM critique-refinement loops.

cs.CL↗

arXiv2024

Pockels Laser Directly Driving Ultrafast Optical Metrology

The invention of the laser unleashed the potential of optical metrology, leading to numerous advancements in modern science and technology. This reliance on lasers, however, also sets a bottleneck for precision optical metrology which is complicated by sophisticated photonic infrastructure required for delicate laser-wave control, leading to limited metrology performance and significant system complexity. Here we make a key step towards resolving this challenge, by demonstrating a Pockels laser with multi-functional capability that advances the optical metrology to a new level. The chip-scale laser exhibits a narrow intrinsic linewidth down to 167 Hz and a broad mode-hop-free tuning range up to 24 GHz. In particular, it offers an unprecedented frequency chirping rate up to 20 EHz/s, and an enormous modulation bandwidth >10 GHz, both orders of magnitude larger than any existing lasers. With this laser, we are able to successfully achieve velocimetry of 40 m/s at a short distance of 0.4 m, with a measurable velocity up to the first cosmic velocity at 1 m away, that is inaccessible by conventional ranging approaches, and distance metrology with a ranging resolution of <2 cm. Moreover, for the first time to the best of our knowledge, we are able to realize a dramatically simplified architecture for laser frequency stabilization, by direct locking the laser to an external reference gas cell without any extra external light control. We successfully achieve a long-term laser stability with a frequency fluctuation of only $\pm$ 6.5 MHz over 60 minutes. The demonstrated Pockels laser combines elegantly high laser coherence with ultrafast frequency reconfigurability and superior multifunctional capability that we envision to have profound impacts on many areas including communication, sensing, autonomous driving, quantum information processing, and beyond.

physics.optics↗

arXiv2024

Harnessing micro-Fabry-Perot reference cavities in photonic integrated circuits

Compact photonic systems that offer high frequency stability and low noise are of increasing importance to applications in precision metrology, quantum computing, communication, and advanced sensing technologies. However, on-chip resonators comprised of dielectrics cannot match the frequency stability and noise characteristics of Fabry-Perot cavities, whose electromagnetic modes live almost entirely in vacuum. In this study, we present a novel strategy to interface micro-fabricated Fabry-Perot cavities with photonic integrated circuits to realize compact, high-performance integrated systems. Using this new integration approach, we demonstrate self-injection locking of an on-chip laser to a milimeter-scale vacuum-gap Fabry-Perot using a circuit interface that transforms the reflected cavity response to enable efficient feedback to the laser. This system achieves a phase noise of -97 dBc/Hz at 10 kHz offset frequency, a fractional frequency stability of 5*10-13 at 10 ms, a 150 Hz 1/pi integral linewidth, and a 35 mHz fundamental linewidth. We also present a complementary integration strategy that utilizes a vertical emission grating coupler and a back-reflection cancellation circuit to realize a fully co-integrated module that effectively redirects the reflected signals and isolates back-reflections with a 10 dB suppression ratio, readily adaptable for on-chip PDH locking. Together, these demonstrations significantly enhance the precision and functionality of RF photonic systems, paving the way for continued advancements in photonic applications.

physics.optics↗

arXiv2024

Investigation of Q degradation in low-loss Si3N4 from heterogeneous laser integration

High-performance, high-volume-manufacturing Si3N4 photonics requires extremely low waveguide losses augmented with heterogeneously integrated lasers for applications beyond traditional markets of high-capacity interconnects. State-of-the-art quality factors (Q) over 200 million at 1550 nm have been shown previously; however, maintaining high Qs throughout laser fabrication has not been shown. Here, Si3N4 resonator intrinsic Qs over 100 million are demonstrated on a fully integrated heterogeneous laser platform. Qi is measured throughout laser processing steps, showing degradation down to 50 million from dry etching, metal evaporation, and ion implant steps, and controllable recovery to over 100 million from annealing at 250C - 350C.

physics.optics↗

arXiv2023

A Novel Approach to Interface High-Q Fabry-Pérot Resonators with Photonic Circuits

The unique benefits of Fabry-Pérot resonators as frequency-stable reference cavities and as an efficient interface between atoms and photons make them an indispensable resource for emerging photonic technologies. To bring these performance benefits to next-generation communications, computation, and timekeeping systems, it will be necessary to develop strategies to integrate compact Fabry-Pérot resonators with photonic integrated circuits. In this paper, we demonstrate a novel reflection cancellation circuit that utilizes a numerically optimized multi-port polarization-splitting grating coupler to efficiently interface high-finesse Fabry-Pérot resonators with a silicon photonic circuit. This circuit interface produces spatial separation of the incident and reflected waves, as required for on-chip Pound-Drever-Hall frequency locking, while also suppressing unwanted back reflections from the Fabry-Pérot resonator. Using inverse design principles, we design and fabricate a polarization-splitting grating coupler that achieves 55% coupling efficiency. This design realizes an insertion loss of 5.8 dB for the circuit interface and more than 9 dB of back reflection suppression, and we demonstrate the versatility of this system by using it to interface several reflective off-chip devices.

physics.optics↗

arXiv2023

An Empirical Study of NetOps Capability of Pre-Trained Large Language Models

Nowadays, the versatile capabilities of Pre-trained Large Language Models (LLMs) have attracted much attention from the industry. However, some vertical domains are more interested in the in-domain capabilities of LLMs. For the Networks domain, we present NetEval, an evaluation set for measuring the comprehensive capabilities of LLMs in Network Operations (NetOps). NetEval is designed for evaluating the commonsense knowledge and inference ability in NetOps in a multi-lingual context. NetEval consists of 5,732 questions about NetOps, covering five different sub-domains of NetOps. With NetEval, we systematically evaluate the NetOps capability of 26 publicly available LLMs. The results show that only GPT-4 can achieve a performance competitive to humans. However, some open models like LLaMA 2 demonstrate significant potential.

cs.CL↗

arXiv2023

Photonic chip-based low noise microwave oscillator

Numerous modern technologies are reliant on the low-phase noise and exquisite timing stability of microwave signals. Substantial progress has been made in the field of microwave photonics, whereby low noise microwave signals are generated by the down-conversion of ultra-stable optical references using a frequency comb. Such systems, however, are constructed with bulk or fiber optics and are difficult to further reduce in size and power consumption. Our work addresses this challenge by leveraging advances in integrated photonics to demonstrate low-noise microwave generation via two-point optical frequency division. Narrow linewidth self-injection locked integrated lasers are stabilized to a miniature Fabry-Pérot cavity, and the frequency gap between the lasers is divided with an efficient dark-soliton frequency comb. The stabilized output of the microcomb is photodetected to produce a microwave signal at 20 GHz with phase noise of -96 dBc/Hz at 100 Hz offset frequency that decreases to -135 dBc/Hz at 10 kHz offset--values which are unprecedented for an integrated photonic system. All photonic components can be heterogeneously integrated on a single chip, providing a significant advance for the application of photonics to high-precision navigation, communication and timing systems.

physics.optics↗

arXiv2023

Engineered zero-dispersion microcombs using CMOS-ready photonics

Normal group velocity dispersion (GVD) microcombs offer high comb line power and high pumping efficiency compared to bright pulse microcombs. The recent demonstration of normal GVD microcombs using CMOS-foundry-produced microresonators is an important step towards scalable production. However, the chromatic dispersion of CMOS devices is large and impairs generation of broadband microcombs. Here, we report the development of a microresonator in which GVD is reduced due to a couple-ring resonator configuration. Operating in the turnkey self-injection-locking mode, the resonator is hybridly integrated with a semiconductor laser pump to produce high-power-efficiency combs spanning a bandwidth of 9.9 nm (1.22 THz) centered at 1560 nm, corresponding to 62 comb lines. Fast, linear optical sampling of the comb waveform is used to observe the rich set of near-zero GVD comb behaviors, including soliton molecules, switching waves (platicons) and their hybrids. Tuning of the 20 GHz repetition rate by electrical actuation enables servo locking to a microwave reference, which simultaneously stabilizes the comb repetition rate, offset frequency and temporal waveform. This hybridly integrated system could be used in coherent communications or for ultra-stable microwave signal generation by two-point optical frequency division.

physics.optics↗

arXiv2023

Three-dimensional integration enables ultra-low-noise, isolator-free Si photonics

While photonic integrated circuits (PICs) are being widely used in applications such as telecommunications and datacenter interconnects, PICs capable of replacing bulk optics and fibers in high-precision, highly-coherent applications will require ultra-low-noise laser sources to be integrated with other photonic components in a compact and robustly aligned format -- that is, on a single chip. Such PICs could offer superior scalability for complex functionalities and volume production, as well as improved stability and reliability over time. However, there are two major issues preventing the realization of such envisioned PICs: the high phase noise of semiconductor lasers, and the difficulty of integrating optical isolators directly on chip. PICs are still considered as inferior solutions in optical systems such as microwave synthesizers, optical gyroscopes and atomic clocks, despite their advantages in size, weight, power consumption and cost (SWaPC). Here, we challenge this convention by introducing three-dimensional (3D) integration in silicon photonics that results in ultra-low-noise, isolator-free PICs. Through multiple monolithic and heterogeneous processing sequences, direct on-chip integration of III-V gain and ultra-low-loss (ULL) silicon nitride (SiN) waveguides with optical loss around 0.5 dB/m are demonstrated. Consequently, the demonstrated PIC enters a new regime, such that an integrated ultra-high-Q cavity reduces the laser noise close to that of fiber lasers. Moreover, the cavity acts as an effective block for any downstream on-chip or off-chip reflection-induced destabilization, thus eliminating the need for optical isolators. We further showcase isolator-free, widely-tunable, low-noise, heterodyne microwave generation using two ultra-low-noise lasers on the same silicon chip.

physics.app-ph↗

arXiv2022

Soliton generation in AlGaAs microresonators at room temperature

Chip-integrated optical frequency combs are attractive optical sources in comb applications requiring high-repetition-rate, low power consumption, or compact size. Spontaneous soliton formation via Kerr parametric oscillation is a promising generation principle in these frequency combs, and has been demonstrated in several material platforms over the past decade. Of these materials, AlGaAs has one of the largest Kerr nonlinearity coefficients allowing low pump threshold comb generation. However, bright soliton generation using this material has only been possible at cryogenic temperature because of the large thermo-optic effect at room temperature, which hinders stable access to the soliton regime. Here, we report self-stabilized single soliton generation in AlGaAs microresonators at room temperature by utilizing a rising soliton step in large free-spectral-range resonators. With sub-milliWatt optical pump power, 1 THz repetition-rate soliton generation is demonstrated. Perfect soliton crystal formation and soliton breather states are also observed. Besides the advantages of large optical nonlinearity, the devices are natural candidates for integration with III-V pump lasers.

physics.optics↗

arXiv2022

Ultralow voltage, High-speed, and Energy-efficient Cryogenic Electro-Optic Modulator

Photonic integrated circuits (PICs) at cryogenic temperatures enable a wide range of applications in scalable classical and quantum systems for computing and sensing. A promising application of cryogenic PICs is to provide optical interconnects by up-converting signals from electrical to optical domain, allowing massive data-transfer from 4 K superconducting (SC) electronics to room temperature environment. Such a solution is central to overcome the major bottleneck in the scalability of cryogenic systems, which currently rely on bulky copper cables that suffer from limited bandwidth, large heat load, and do not show any scalability path. A key element for realizing a cryogenic-to-room temperature optical interconnect is a high-speed electro-optic (EO) modulator operating at 4 K with operation voltage at mV scale, compatible with SC electronics. Although several cryogenic EO modulators have been demonstrated, their driving voltages are significantly large compared to the mV scale voltage required for SC circuits. Here, we demonstrate a cryogenic modulator with ~10 mV peak-to-peak driving voltage and gigabits/sec data rate, with ultra-low electric and optical energy consumptions of ~10.4 atto-joules/bit and ~213 femto-joules/bit, respectively. We achieve this record performance by designing a compact optical ring resonator modulator in a heterogeneous InP-on-Silicon platform, where we optimize a multi-quantum well layer of InAIGaAs to achieve a strong EO effect at 4 K. Unlike other semiconductors such as silicon, our platform benefits from the high-carrier mobility and minimal free carrier freezing of III-V compounds at low temperatures, with moderate doping level and low loss (intrinsic resonator Q~272,000). These modulators can pave the path for complex cryogenic photonic functionalities and massive data transmission between cryogenic and room-temperature electronics.

physics.optics↗

arXiv2022

Chip-Based Laser with 1 Hertz Integrated Linewidth

Lasers with hertz-level linewidths on timescales up to seconds are critical for precision metrology, timekeeping, and manipulation of quantum systems. Such frequency stability typically relies on bulk-optic lasers and reference cavities, where increased size is leveraged to improve noise performance, but with the trade-off of cost, hand assembly, and limited application environments. On the other hand, planar waveguide lasers and cavities exploit the benefits of CMOS scalability but are fundamentally limited from achieving hertz-level linewidths at longer times by stochastic noise and thermal sensitivity inherent to the waveguide medium. These physical limits have inhibited the development of compact laser systems with frequency noise required for portable optical clocks that have performance well beyond conventional microwave counterparts. In this work, we break this paradigm to demonstrate a compact, high-coherence laser system at 1548 nm with a 1 s integrated linewidth of 1.1 Hz and fractional frequency instability less than 10$^{-14}$ from 1 ms to 1 s. The frequency noise at 1 Hz offset is suppressed by 11 orders of magnitude from that of the free-running diode laser down to the cavity thermal noise limit near 1 Hz$^2$/Hz, decreasing to 10$^{-3}$ Hz$^2$/Hz at 4 kHz offset. This low noise performance leverages wafer-scale integrated lasers together with an 8 mL vacuum-gap cavity that employs micro-fabricated mirrors with sub-angstrom roughness to yield an optical $Q$ of 11.8 billion. Significantly, all the critical components are lithographically defined on planar substrates and hold the potential for parallel high-volume manufacturing. Consequently, this work provides an important advance towards compact lasers with hertz-level linewidths for applications such as portable optical clocks, low-noise RF photonic oscillators, and related communication and navigation systems.

physics.optics↗