Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
16works
0followers
15topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

16 published item(s)

preprint2026arXiv

AMAP Agentic Planning Technical Report

We present STAgent, an agentic large language model tailored for spatio-temporal understanding, designed to solve complex tasks such as constrained point-of-interest discovery and itinerary planning. STAgent is a specialized model capable of interacting with ten distinct tools within spatio-temporal scenarios, enabling it to explore, verify, and refine intermediate steps during complex reasoning. Notably, STAgent effectively preserves its general capabilities. We empower STAgent with these capabilities through three key contributions: (1) a stable tool environment that supports over ten domain-specific tools, enabling asynchronous rollout and training; (2) a hierarchical data curation framework that identifies high-quality data like a needle in a haystack, curating high-quality queries by retaining less than 1\% of the raw data, emphasizing both diversity and difficulty; and (3) a cascaded training recipe that starts with a seed SFT stage acting as a guardian to measure query difficulty, followed by a second SFT stage fine-tuned on queries with high certainty, and an ultimate RL stage that leverages data of low certainty. Initialized with Qwen3-30B-A3B to establish a strong SFT foundation and leverage insights into sample difficulty, STAgent yields promising performance on TravelBench while maintaining its general capabilities across a wide range of general benchmarks, thereby demonstrating the effectiveness of our proposed agentic model.

preprint2026arXiv

TurboGR: An Accelerated Training System for Large-Scale Generative Recommendation

Generative recommendation (GR) has emerged as a promising paradigm that replaces fragmented, scenario-specific architectures with unified Transformer-based models, exhibiting scaling-law behavior where recommendation quality improves systematically with increased model capacity and training data. However, deploying GR at scale on Ascend NPUs faces fundamental system-level challenges. These challenges are further exacerbated on Ascend NPUs due to the absence of high-performance implementations for jagged operators and the architectural mismatch between irregular sparse primitives and NPU's dense-computation-optimized design. In this paper, we present \model, an Ascend-affinity training system for generative recommendation that systematically addresses these bottlenecks through three core innovations: (i) Ascend-affinity jagged acceleration, including fusion operators that eliminate padding redundancy and dynamic load balancing that reduces inter-device imbalance from 47\% to 2.4\%; (ii) distributed communication optimization, comprising hierarchical sparse parallelism, semi-asynchronous training with proven convergence guarantees, and fine-grained pipeline orchestration that sustains 94\% NPU utilization; and (iii) negative sampling optimization via asynchronous offloading, jaggedness-aware FP16 quantization, and intra-batch logit sharing that expand the effective negative space without additional embedding lookups. Evaluated on the KuaiRand-27K dataset, \model supports training at up to 0.2B parameters and achieves 54.71\% MFU with near-linear scalability (0.97).

preprint2026arXiv

Unveiling the Shortwave Absorption Spectra of Alumina Aerosols: Implications for Solar Radiation Modification

Alumina is proposed for Stratospheric Aerosol Injection (SAI)-based solar radiation modification due to its presumed ability to scatter sunlight strongly while absorbing weakly. Alumina is assigned negligible solar shortwave absorption in climate models; this assumption is not validated owing to technological challenges in quantifying its weak absorption signals. We report alumina's shortwave imaginary refractive index $k$, a determinant of its absorption strength, using sensitive in situ photoacoustic spectrometry, finding values ranging from $1.4 \times 10^{-4}$ to $1.2 \times 10^{-3}$. Particle-scale electron energy-loss spectroscopy provided independent validation and revealed that the non-ideal absorption arises from oxygen vacancy defects in the alumina's crystal structure. Aerosol chemistry climate model simulations to evaluate shortwave absorption radiative effects revealed insignificant impacts on radiative forcing and stratospheric warming. Our findings indicate that alumina's shortwave absorption, previously reported as a source of uncertainty, is unlikely to affect SAI impact calculations.

preprint2023arXiv

Efficient and scalable scheme for overcoming the pulse energy bottleneck of single-cycle laser sources

We propose a novel scheme called advanced dual-chirped optical parametric amplification (DC-OPA) that employs two kinds of nonlinear crystals (BiB$_3$O$_6$ and MgO-doped lithium niobate) to overcome the bottleneck of pulse energy scalability for single-cycle mid-infrared (MIR) laser pulses. In experiments, the advanced DC-OPA scheme achieved carrier-to-envelope phase-stable MIR laser pulses for a bandwidth of over one octave (1.4-3.1 $μ$m) with an output pulse energy of 53 mJ. The pulse duration was compressed to 8.58 fs, which corresponds to 1.05 cycles with a central wavelength of 2.44 $μ$m and peak power of 6 TW. To our knowledge, the obtained values for the pulse energy and peak power are the highest achieved for optical parametric amplification of single-cycle MIR laser pulses. Thanks to the energy scalability of the advanced DC-OPA scheme, it is potentially applicable to multi-TW sub-cycle laser pulses.

preprint2022arXiv

100-mJ class, sub-two-cycle, carrier-envelope phase-stable dual-chirped optical parametric amplification

Based on the dual-chirped optical parametric amplification and type-I BiB$_3$O$_6$(BiBO) crystals, the generation of $>$100 mJ, 10.4 fs, 10 Hz, carrier-to-envelope phase (CEP)-stable laser pulses, which are centered at 1.7 $μ$m, is demonstrated; it produces a peak power of 10 TW. CEP-dependent high harmonic generation is implemented to confirm the sub-two-cycle pulse duration and CEP stabilization of infrared (IR) laser pulses. As far as we know, the obtained pulse energy and peak power represent the highest values for sub-two-cycle CEP-stable IR optical parametric amplification. Additionally, the prospects of achieving high-energy water window isolated attosecond pulses via our developed laser source are discussed.

preprint2022arXiv

Document-level Relation Extraction with Context Guided Mention Integration and Inter-pair Reasoning

Document-level Relation Extraction (DRE) aims to recognize the relations between two entities. The entity may correspond to multiple mentions that span beyond sentence boundary. Few previous studies have investigated the mention integration, which may be problematic because coreferential mentions do not equally contribute to a specific relation. Moreover, prior efforts mainly focus on reasoning at entity-level rather than capturing the global interactions between entity pairs. In this paper, we propose two novel techniques, Context Guided Mention Integration and Inter-pair Reasoning (CGM2IR), to improve the DRE. Instead of simply applying average pooling, the contexts are utilized to guide the integration of coreferential mentions in a weighted sum manner. Additionally, inter-pair reasoning executes an iterative algorithm on the entity pair graph, so as to model the interdependency of relations. We evaluate our CGM2IR model on three widely used benchmark datasets, namely DocRED, CDR, and GDA. Experimental results show that our model outperforms previous state-of-the-art models.

preprint2022arXiv

Feature-Aligned Video Raindrop Removal with Temporal Constraints

Existing adherent raindrop removal methods focus on the detection of the raindrop locations, and then use inpainting techniques or generative networks to recover the background behind raindrops. Yet, as adherent raindrops are diverse in sizes and appearances, the detection is challenging for both single image and video. Moreover, unlike rain streaks, adherent raindrops tend to cover the same area in several frames. Addressing these problems, our method employs a two-stage video-based raindrop removal method. The first stage is the single image module, which generates initial clean results. The second stage is the multiple frame module, which further refines the initial results using temporal constraints, namely, by utilizing multiple input frames in our process and applying temporal consistency between adjacent output frames. Our single image module employs a raindrop removal network to generate initial raindrop removal results, and create a mask representing the differences between the input and initial output. Once the masks and initial results for consecutive frames are obtained, our multiple-frame module aligns the frames in both the image and feature levels and then obtains the clean background. Our method initially employs optical flow to align the frames, and then utilizes deformable convolution layers further to achieve feature-level frame alignment. To remove small raindrops and recover correct backgrounds, a target frame is predicted from adjacent frames. A series of unsupervised losses are proposed so that our second stage, which is the video raindrop removal module, can self-learn from video data without ground truths. Experimental results on real videos demonstrate the state-of-art performance of our method both quantitatively and qualitatively.

preprint2022arXiv

Hydrodynamics for one-dimensional ASEP in contact with a class of reservoirs

We study the hydrodynamic behaviour of the asymmetric simple exclusion process on the lattice of size $n$. In the bulk, the exclusion dynamics performs rightward flux. At the boundaries, the dynamics is attached to reservoirs. We investigate two types of reservoirs: (1) the reservoirs that are weakened by $n^θ$ for some $θ<0$ and (2) the reservoirs that create particles only at the right boundary and annihilate particles only at the left boundary. We prove that the spatial density of particles, under the hyperbolic time scale, evolves with the entropy solution to a scalar conservation law on $[0,1]$ with boundary conditions. The boundary conditions are characterised by the boundary traces at $x=0$ and $x=1$ which take values from $\{0,1\}$.

preprint2022arXiv

Parallel Network with Channel Attention and Post-Processing for Carotid Arteries Vulnerable Plaque Segmentation in Ultrasound Images

Carotid arteries vulnerable plaques are a crucial factor in the screening of atherosclerosis by ultrasound technique. However, the plaques are contaminated by various noises such as artifact, speckle noise, and manual segmentation may be time-consuming. This paper proposes an automatic convolutional neural network (CNN) method for plaque segmentation in carotid ultrasound images using a small dataset. First, a parallel network with three independent scale decoders is utilized as our base segmentation network, pyramid dilation convolutions are used to enlarge receptive fields in the three segmentation sub-networks. Subsequently, the three decoders are merged to be rectified in channels by SENet. Thirdly, in test stage, the initially segmented plaque is refined by the max contour morphology post-processing to obtain the final plaque. Moreover, three loss function Dice loss, SSIM loss and cross-entropy loss are compared to segment plaques. Test results show that the proposed method with dice loss function yields a Dice value of 0.820, an IoU of 0.701, Acc of 0.969, and modified Hausdorff distance (MHD) of 1.43 for 30 vulnerable cases of plaques, it outperforms some of the conventional CNN-based methods on these metrics. Additionally, we apply an ablation experiment to show the validity of each proposed module. Our study provides some reference for similar researches and may be useful in actual applications for plaque segmentation of ultrasound carotid arteries.

preprint2022arXiv

Quasi-static limit for the asymmetric simple exclusion

We study the one-dimensional asymmetric simple exclusion process on the lattice $\{1, \dots,N\}$ with creation/annihilation at the boundaries. The boundary rates are time dependent and change on a slow time scale $N^{-a}$ with $a>0$. We prove that at the time scale $N^{1+a}$ the system evolves quasi-statically with a macroscopic density profile given by the entropy solution of the stationary Burgers equation with boundary densities changing in time, determined by the corresponding microscopic boundary rates. We consider two different types of boundary rates: the &#34;Liggett boundaries&#34; that correspond to the projection of the infinite dynamics, and the reversible boundaries, that correspond to the contact with particle reservoirs in equilibrium. The proof is based on the control of the Lax boundary entropy--entropy flux pairs and a coupling argument.

preprint2021arXiv

A new volatility model: GQARCH-Itô model

Volatility asymmetry is a hot topic in high-frequency financial market. In this paper, we propose a new econometric model, which could describe volatility asymmetry based on high-frequency historical data and low-frequency historical data. After providing the quasi-maximum likelihood estimators for the parameters, we establish their asymptotic properties. We also conduct a series of simulation studies to check the finite sample performance and volatility forecasting performance of the proposed methodologies. And an empirical application is demonstrated that the new model has stronger volatility prediction power than GARCH-Itô model in the literature.

preprint2021arXiv

Position-Aware Tagging for Aspect Sentiment Triplet Extraction

Aspect Sentiment Triplet Extraction (ASTE) is the task of extracting the triplets of target entities, their associated sentiment, and opinion spans explaining the reason for the sentiment. Existing research efforts mostly solve this problem using pipeline approaches, which break the triplet extraction process into several stages. Our observation is that the three elements within a triplet are highly related to each other, and this motivates us to build a joint model to extract such triplets using a sequence tagging approach. However, how to effectively design a tagging approach to extract the triplets that can capture the rich interactions among the elements is a challenging research question. In this work, we propose the first end-to-end model with a novel position-aware tagging scheme that is capable of jointly extracting the triplets. Our experimental results on several existing datasets show that jointly capturing elements in the triplet using our approach leads to improved performance over the existing approaches. We also conducted extensive experiments to investigate the model effectiveness and robustness.

preprint2020arXiv

Hyperbolic scaling limit of non-equilibrium fluctuations for a weakly anharmonic chain

We consider a chain of $n$ coupled oscillators placed on a one-dimensional lattice with periodic boundary conditions. The interaction between particles is determined by a weakly anharmonic potential $V_n = r^2/2 + σ_nU(r)$, where $U$ has bounded second derivative and $σ_n$ vanishes as $n \to \infty$. The dynamics is perturbed by noises acting only on the positions, such that the total momentum and length are the only conserved quantities. With relative entropy technique, we prove for dynamics out of equilibrium that, if $σ_n$ decays sufficiently fast, the fluctuation field of the conserved quantities converges in law to a linear $p$-system in the hyperbolic space-time scaling limit. The transition speed is spatially homogeneous due to the vanishing anharmonicity. We also present a quantitative bound for the speed of convergence to the corresponding hydrodynamic limit.

preprint2020arXiv

Optimization of a multi-TW few-cycle 1.7-$μ$m source based on Type-I BBO dual-chirped optical parametric amplification

This paper presents the optimization of a dual-chirped optical parametric amplification (DC-OPA) scheme for producing an ultrafast intense infrared (IR) pulse. By employing a total energy of 0.77 J Ti:sapphire pump laser and type-I BBO crystals, an IR pulse energy at the center wavelength of 1.7 $μ$m exceeded 0.1 J using the optimized DC-OPA. By adjusting the injected seed spectrum and prism pair compressor with a gross throughput of over 70 \%, the 1.7-$μ$m pulse was compressed to 31 fs, which resulted in a peak power of up to 2.3 TW. Based on the demonstration of the BBO type-I DC-OPA, we propose a novel OPA scheme called the $dual~pump$ DC-OPA for producing a high-energy IR pulse with a two-cycle duration.

preprint2019arXiv

Equilibrium fluctuation for an anharmonic chain with boundary conditions in the Euler scaling limit

We study the evolution in equilibrium of the fluctuations for the conserved quantities of a chain of anharmonic oscillators in the hyperbolic space-time scaling. Boundary conditions are determined by applying a constant tension at one side, while the position of the other side is kept fixed. The Hamiltonian dynamics is perturbed by random terms conservative of such quantities. We prove that these fluctuations evolve macroscopically following the linearized Euler equations with the corresponding boundary conditions, even in some time scales larger than the hyperbolic one.