arXiv subjects

Yuan Chai

Publications and source records attributed to Yuan Chai.

11 recordsLinked to original sources

arXiv2026

RFS-UNet: Decoder-Conditioned High-Resolution Skip Recalibration for Bone-Selective DRR Synthesis

Bone-selective digitally reconstructed radiograph (DRR) synthesis depends on high-resolution encoder detail, yet static skips cannot condition reuse on the evolving decoder representation. We ask whether decoder state adds useful information beyond encoder-only self-recalibration for high-resolution skip reuse. RFS-UNet uses pooled encoder and aligned decoder statistics for bounded residual channel recalibration at the 512^2 and 256^2 skips, leaving the backbone unchanged. In the matched seed-2026 comparison isolating decoder conditioning, RFS raises validation PSNR by 0.254 dB over Self-RFS. Across three seeds, locked-test PSNR rises from 33.225+/-0.048 to 33.537+/-0.128 dB; RFS lowers MAE in 179/200 held-out CT cases and reduces mean MAE by 3.91%. It adds 0.117% parameters and 1.169% counted Conv2d operations. These results support decoder state as a useful conditioning signal for high-resolution feature reuse in controlled paired projection synthesis.

cs.CV↗

arXiv2026

Twisted Kazhdan-Lusztig conjecture for $p$-adic general linear group

We use enhanced Langlands parameters to obtain a classification for irreducible representations of twisted $p$-adic general linear groups in unramified principal series. We give the definition of standard representations and prove the twisted Kazhdan-Lusztig conjecture for the multiplicities in the Grothendieck group. We mainly follow Lusztig's work in the connected case using graded Hecke algebra. We show that the parametrization is compatible with the Whittaker-normalized one.

math.RT↗

arXiv2026

Opportunistic Bone-Loss Screening from Routine Knee Radiographs Using a Multi-Task Deep Learning Framework with Sensitivity-Constrained Threshold Optimization

Background: Osteoporosis and osteopenia are often undiagnosed until fragility fractures occur. Dual-energy X-ray absorptiometry (DXA) is the reference standard for bone mineral density (BMD) assessment, but access remains limited. Knee radiographs are obtained at high volume for osteoarthritis evaluation and may offer an opportunity for opportunistic bone-loss screening. Objective: To develop and evaluate a multi-task deep learning system for opportunistic bone-loss screening from routine knee radiographs without additional imaging or patient visits. Methods: We developed STR-Net, a multi-task framework for single-channel grayscale knee radiographs. The model includes a shared backbone, global average pooling feature aggregation, a shared neck, and a task-aware representation routing module connected to three task-specific heads: binary screening (Normal vs. Bone Loss), severity sub-classification (Osteopenia vs. Osteoporosis), and weakly coupled T-score regression with optional clinical variables. A sensitivity-constrained threshold optimization strategy (minimum sensitivity >= 0.86) was applied. The dataset included 1,570 knee radiographs, split at the patient level into training (n=1,120), validation (n=226), and test (n=224) sets. Results: On the held-out test set, STR-Net achieved an AUROC of 0.933, sensitivity of 0.904, specificity of 0.773, and AUPRC of 0.956 for binary screening. Severity sub-classification achieved an AUROC of 0.898. The T-score regression branch showed a Pearson correlation of 0.801 with DXA-measured T-scores in a pilot subset (n=31), with MAE of 0.279 and RMSE of 0.347. Conclusions: STR-Net enables single-pass bone-loss screening, severity stratification, and quantitative T-score estimation from routine knee radiographs. Prospective clinical validation is needed before deployment.

cs.CV↗

arXiv2026

AGE-Net: Spectral--Spatial Fusion and Anatomical Graph Reasoning with Evidential Ordinal Regression for Knee Osteoarthritis Grading

Automated Kellgren--Lawrence (KL) grading from knee radiographs is challenging due to subtle structural changes, long-range anatomical dependencies, and ambiguity near grade boundaries. We propose AGE-Net, a ConvNeXt-based framework that integrates Spectral--Spatial Fusion (SSF), Anatomical Graph Reasoning (AGR), and Differential Refinement (DFR). To capture predictive uncertainty and preserve label ordinality, AGE-Net employs a Normal-Inverse-Gamma (NIG) evidential regression head and a pairwise ordinal ranking constraint. On a knee KL dataset, AGE-Net achieves a quadratic weighted kappa (QWK) of 0.9017 +/- 0.0045 and a mean squared error (MSE) of 0.2349 +/- 0.0028 over three random seeds, outperforming strong CNN baselines and showing consistent gains in ablation studies. We further outline evaluations of uncertainty quality, robustness, and explainability, with additional experimental figures to be included in the full manuscript.

cs.CV↗

arXiv2025

UniDex: Rethinking Search Inverted Indexing with Unified Semantic Modeling

Inverted indexing has traditionally been a cornerstone of modern search systems, leveraging exact term matches to determine relevance between queries and documents. However, this term-based approach often emphasizes surface-level token overlap, limiting the system's generalization capabilities and retrieval effectiveness. To address these challenges, we propose UniDex, a novel model-based method that employs unified semantic modeling to revolutionize inverted indexing. UniDex replaces complex manual designs with a streamlined architecture, enhancing semantic generalization while reducing maintenance overhead. Our approach involves two key components: UniTouch, which maps queries and documents into semantic IDs for improved retrieval, and UniRank, which employs semantic matching to rank results effectively. Through large-scale industrial datasets and real-world online traffic assessments, we demonstrate that UniDex significantly improves retrieval capabilities, marking a paradigm shift from term-based to model-based indexing. Our deployment within Kuaishou's short-video search systems further validates UniDex's practical effectiveness, serving hundreds of millions of active users efficiently.

cs.IR↗

arXiv2025

Laser optothermal nanobomb for efficient flattening of nanobubbles in van der Waals materials

Nanobubbles are typical nanodefects commonly existing in two-dimensional (2D) van der Waals materials such as transition metal dioxides, especially after their transfer from growth substrate to target substrates. These nanobubbles, though tiny, may significantly alter the local electric, optoelectronic, thermal, or mechanical properties of 2D materials and therefore are rather detrimental to the constructed devices. However, there is no post-processing method so far that can effectively eliminate nanobubbles in 2D materials after their fabrication and transfer, which has been a major obstacle in the development of 2D material based devices. Here, we propose a principle, called laser optothermal nanobomb (LOTB), that can effectively flatten nanobubbles in 2D materials through a dynamic process of optothermally induced phase transition and stress-pulling effect in nanobubbles. Operation of LOTB on monolayer molybdenum disulfide (1L-MoS2) films shows that the surface roughness can be reduced by more than 70% on a time scale of ~50 ms, without damage to the intrinsic property of 1L-MoS2 as validated by micro-nano photoluminescence and Raman spectroscopy. Moreover, a dual-beam cascaded LOTB and a multi-shot LOTB strategies are proposed to increase the flattened area and processing effect, showing the potential of LOTB for fast nanodefect repairing in the mass production of van der Waals materials and devices.

physics.optics↗

arXiv2024

RecFlow: An Industrial Full Flow Recommendation Dataset

Industrial recommendation systems (RS) rely on the multi-stage pipeline to balance effectiveness and efficiency when delivering items from a vast corpus to users. Existing RS benchmark datasets primarily focus on the exposure space, where novel RS algorithms are trained and evaluated. However, when these algorithms transition to real world industrial RS, they face a critical challenge of handling unexposed items which are a significantly larger space than the exposed one. This discrepancy profoundly impacts their practical performance. Additionally, these algorithms often overlook the intricate interplay between multiple RS stages, resulting in suboptimal overall system performance. To address this issue, we introduce RecFlow, an industrial full flow recommendation dataset designed to bridge the gap between offline RS benchmarks and the real online environment. Unlike existing datasets, RecFlow includes samples not only from the exposure space but also unexposed items filtered at each stage of the RS funnel. Our dataset comprises 38M interactions from 42K users across nearly 9M items with additional 1.9B stage samples collected from 9.3M online requests over 37 days and spanning 6 stages. Leveraging the RecFlow dataset, we conduct courageous exploration experiments, showcasing its potential in designing new algorithms to enhance effectiveness by incorporating stage-specific samples. Some of these algorithms have already been deployed online, consistently yielding significant gains. We propose RecFlow as the first comprehensive benchmark dataset for the RS community, supporting research on designing algorithms at any stage, study of selection bias, debiased algorithms, multi-stage consistency and optimality, multi-task recommendation, and user behavior modeling. The RecFlow dataset, along with the corresponding source code, is available at https://github.com/RecFlow-ICLR/RecFlow.

cs.IR↗

arXiv2022

Cross-Lingual Ability of Multilingual Masked Language Models: A Study of Language Structure

Multilingual pre-trained language models, such as mBERT and XLM-R, have shown impressive cross-lingual ability. Surprisingly, both of them use multilingual masked language model (MLM) without any cross-lingual supervision or aligned data. Despite the encouraging results, we still lack a clear understanding of why cross-lingual ability could emerge from multilingual MLM. In our work, we argue that cross-language ability comes from the commonality between languages. Specifically, we study three language properties: constituent order, composition and word co-occurrence. First, we create an artificial language by modifying property in source language. Then we study the contribution of modified property through the change of cross-language transfer results on target language. We conduct experiments on six languages and two cross-lingual NLP tasks (textual entailment, sentence retrieval). Our main conclusion is that the contribution of constituent order and word co-occurrence is limited, while the composition is more crucial to the success of cross-linguistic transfer.

cs.CL↗

arXiv2021

CogTree: Cognition Tree Loss for Unbiased Scene Graph Generation

Scene graphs are semantic abstraction of images that encourage visual understanding and reasoning. However, the performance of Scene Graph Generation (SGG) is unsatisfactory when faced with biased data in real-world scenarios. Conventional debiasing research mainly studies from the view of balancing data distribution or learning unbiased models and representations, ignoring the correlations among the biased classes. In this work, we analyze this problem from a novel cognition perspective: automatically building a hierarchical cognitive structure from the biased predictions and navigating that hierarchy to locate the relationships, making the tail relationships receive more attention in a coarse-to-fine mode. To this end, we propose a novel debiasing Cognition Tree (CogTree) loss for unbiased SGG. We first build a cognitive structure CogTree to organize the relationships based on the prediction of a biased SGG model. The CogTree distinguishes remarkably different relationships at first and then focuses on a small portion of easily confused ones. Then, we propose a debiasing loss specially for this cognitive structure, which supports coarse-to-fine distinction for the correct relationships. The loss is model-agnostic and consistently boosting the performance of several state-of-the-art models. The code is available at: https://github.com/CYVincent/Scene-Graph-Transformer-CogTree.

cs.CV↗

arXiv2021

Damage accumulation during high temperature fatigue of Ti/SiC$_f$ metal matrix composites under different stress amplitudes

The damage mechanisms and load redistribution of high strength TC17 titanium alloy/unidirectional SiC fibre composite (fibre diameter = 100 $μ$m) under high temperature (350 °C) fatigue cycling have been investigated in situ using synchrotron X-ray computed tomography (CT) and X-ray diffraction (XRD) for high cycle fatigue (HCF) under different stress amplitudes. The three-dimensional morphology of the crack and fibre fractures has been mapped by CT. During stable growth, matrix cracking dominates with the crack deflecting (by 50-100 $μ$m in height) when bypassing bridging fibres. A small number of bridging fibres have fractured close to the matrix crack plane especially under relatively high stress amplitude cycling. Loading to the peak stress led to rapid crack growth accompanied by a burst of fibre fractures. Many of the fibre fractures occurred 50-300 $μ$m from the matrix crack plane during rapid growth, in contrast to that in the stable growth stage, leading to extensive fibre pull-out on the fracture surface. The changes in fibre loading, interfacial stress, and the extent of fibre-matrix debonding in the vicinity of the crack have been mapped for the fatigue cycle and after the rapid growth by high spatial resolution XRD. The fibre/matrix interfacial sliding extends up to 600 $μ$m (in the stable growth zone) or 700 $μ$m (in the rapid growth zone) either side of the crack plane. The direction of interfacial shear stress reverses with the loading cycle, with the maximum frictional sliding stress reaching ~55 MPa in both the stable growth and rapid growth regimes.

cond-mat.mtrl-sci↗

arXiv2020

Gait Graph Optimization: Generate Variable Gaits from One Base Gait for Lower-limb Rehabilitation Exoskeleton Robots

The most concentrated application of lower-limb rehabilitation exoskeleton (LLE) robot is that it can help paraplegics "re-walk". However, "walking" in daily life is more than just walking on flat ground with fixed gait. This paper focuses on variable gaits generation for LLE robot to adapt complex walking environment. Different from traditional gaits generator for biped robot, the generated gaits for LLEs should be comfortable to patients. Inspired by the pose graph optimization algorithm in SLAM, we propose a graph-based gait generation algorithm called gait graph optimization (GGO) to generate variable, functional and comfortable gaits from one base gait collected from healthy individuals to adapt the walking environment. Variants of walking problem, e.g., stride adjustment, obstacle avoidance, and stair ascent and descent, help verify the proposed approach in simulation and experimentation. We open source our implementation.

cs.RO↗