arXiv subjects

Yanlin Chen

Publications and source records attributed to Yanlin Chen.

At least 19 recordsLinked to original sources

arXiv2026

"World Knowledge" in the Weights: Reading Concept Circuits of Vision Transformers

Vision transformers (ViTs) have achieved remarkable generalization across visual domains, yet little is known about how they internally represent the structure of the world. To address this gap, we use Cross-Layer Transcoders (CLTs) to read concept circuits from ViTs: directed graphs whose nodes correspond to sparse, interpretable concepts and edges capture concept interactions across layers. Our method yields two complementary views of model behavior. The global concept circuit is input-invariant and can be recovered directly from learned cross-layer weights, exposing the reusable "world knowledge" encoded in the model. The instance concept circuit is input-dependent and identifies the concepts and pathways actually used for a specific prediction, enabling faithful example-level explanations. We demonstrate the utility of concept circuits in three ways: (1) Automatic spurious correlation discovery: leveraging the statistics of our global concept circuits to identify shortcut dependencies within the model. (2) Spurious correlation removal: intervening on the instance concept circuit to steer the model towards correct predictions. Empirical results show that our method outperforms existing counterparts by 11.0% on the Waterbird dataset. (3) Model comparison: contrasting the global concept circuits of different foundation models (e.g., CLIP vs. DINO) to reveal how supervision paradigms shape representational structure. Our code is available at https://github.com/deep-real/VisionCLT

cs.CV↗

arXiv2026

AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems

Improving an industrial recommender is an iterative research-and-engineering process rather than a direct path from idea to deployment. In \textbf{DASHEN, NetEase's gaming-community app}, algorithm engineers typically identify promising directions from research papers, technical reports, and prior production experiments; reproduce or adapt the underlying methods; implement them in the production codebase; and evaluate the resulting models through training and offline experiments. Promising candidates are then advanced to online A/B tests, and those demonstrating robust gains are submitted to Launch Review---the internal gate for full-traffic rollout. Large language models (LLMs) can assist with individual stages of this workflow, but the overall process remains human-dependent without a harness that can reliably coordinate them across long-running, often multi-day experimental cycles. We present \textbf{AutoLR}, initially built as \textbf{Auto Launch Review} and later extended upstream into an autonomous research-to-launch harness. AutoLR combines three system mechanisms: a \textbf{multi-expert council} that debates and adversarially reviews proposals; a \textbf{deterministic evidence-weighted exploration--exploitation selector} that allocates a limited trial budget across candidate directions and uses Council reranking; and a layered knowledge system that combines external research, production-system knowledge, and DASHEN-specific domain knowledge---such as game communities, player characteristics, and content-interaction patterns---with posterior evidence from configurations, patches, logs, failures, and offline outcomes. LLM agents perform semantic reasoning and code generation, while deterministic controllers retain authority over execution, metric extraction, guardrails, and persistent state transitions.

cs.AI↗

arXiv2026

An Improved Quantum Algorithm for 3-Tuple Lattice Sieving

The assumed hardness of the Shortest Vector Problem in high-dimensional lattices is one of the cornerstones of post-quantum cryptography. The fastest known heuristic attacks on SVP are via so-called sieving methods. While these still take exponential time in the dimension $d$, they are significantly faster than non-heuristic approaches and their heuristic assumptions are verified by extensive experiments. $k$-Tuple sieving is an iterative method where each iteration takes as input a large number of lattice vectors of a certain norm, and produces an equal number of lattice vectors of slightly smaller norm, by taking sums and differences of $k$ of the input vectors. Iterating these ''sieving steps'' sufficiently many times produces a short lattice vector. The fastest attacks (both classical and quantum) are for $k=2$, but taking larger $k$ reduces the amount of memory required for the attack. In this paper we improve the quantum time complexity of 3-tuple sieving from $2^{0.3098 d}$ to $2^{0.2846 d}$, using a two-level amplitude amplification aided by a preprocessing step that associates the given lattice vectors with nearby ''center points'' to focus the search on the neighborhoods of these center points. Our algorithm uses $2^{0.1887d}$ classical bits and QCRAM bits, and $2^{o(d)}$ qubits. This is the fastest known quantum algorithm for SVP when total memory is limited to $2^{0.1887d}$.

quant-ph↗

arXiv2026

Propeller-Assisted Robust 3D Hopping Robot with Hierarchical Force Allocation

Monopedal hopping robots are conceptually simple but highly dynamic and inherently unstable. Achieving robust 3D hopping is still difficult because ground reaction forces are available only during the short stance phase, while the robot is underactuated in flight. A key unresolved issue is how to improve flight-phase control authority. Propeller assistance provides a promising solution, but it requires careful coordination of leg-generated contact forces and propeller thrusts across stance and flight. This paper presents Pro-OMEGA2, a propeller-assisted 3D monopedal hopping robot with an active 3-RSR parallel leg and a trunk-mounted tri-rotor for auxiliary attitude regulation. To address the force coordination challenge, we propose a Hierarchical Force Allocation (HFA) framework based on a single rigid body (SRB) model. The leg generates the main stance contact wrench, while the tri-rotor provides auxiliary attitude regulation, compensating the residual attitude moment in stance and maintaining attitude during flight. Real-robot experiments in indoor and outdoor scenarios demonstrate sustained 3D hopping, including terrain transitions and impulsive push recovery, validating robustness under unmodeled contact and external disturbances.

cs.RO↗

arXiv2026

Inside the Visual Mind: Neuroscience-Motivated Concept Circuits for Interpreting and Steering Vision Transformers

Despite high accuracy, Vision Transformer (ViT) predictions can be driven by spurious cues, raising the need to understand their inner workings before safe deployment. Sparse autoencoders (SAEs) provide a promising lens for decomposing model representations into human-interpretable concepts, yet adapting SAE-based interpretation to ViTs remains challenging due to limited control over concept coverage and subjective, non-scalable feature interpretation. To fill the gaps, motivated by neuroscience-inspired principles, we propose ViSAE, a mechanistic interpretability toolbox for understanding ViT inner workings through concept circuits. ViSAE consists of three components: (1) A probing suite with 64K images and a 16K visually grounded concept vocabulary, improving concept coverage efficiency by 20x over ImageNet and interpretation accuracy by 28.7% over existing concept sets. (2) Top-down concept reading and Bottom-up circuit tracing algorithms that automatically recover ViT inner workings via concept circuits. (3) Applications for auditing and steering ViT behavior. Through concept editing, ViSAE improves the worst-group accuracy on WaterBirds by 48.2%, outperforming existing methods by 23.8%. Our data and code: https://github.com/deep-real/ViSAE.

cs.CV↗

arXiv2025

Physics-Embedded Gaussian Process for Traffic State Estimation

Traffic state estimation (TSE) becomes challenging when probe-vehicle penetration is low and observations are spatially sparse. Pure data-driven methods lack physical explanations and have poor generalization when observed data is sparse. In contrast, physical models have difficulty integrating uncertainties and capturing the real complexity of traffic. To bridge this gap, recent studies have explored combining them by embedding physical structure into Gaussian process. These approaches typically introduce the governing equations as soft constraints through pseudo-observations, enabling the integration of model structure within a variational framework. However, these methods rely heavily on penalty tuning and lack principled uncertainty calibration, which makes them sensitive to model mis-specification. In this work, we address these limitations by presenting a novel Physics-Embedded Gaussian Process (PEGP), designed to integrate domain knowledge with data-driven methods in traffic state estimation. Specifically, we design two multi-output kernels informed by classic traffic flow models, constructed via the explicit application of the linearized differential operator. Experiments on HighD, NGSIM show consistent improvements over non-physics baselines. PEGP-ARZ proves more reliable under sparse observation, while PEGP-LWR achieves lower errors with denser observation. Ablation study further reveals that PEGP-ARZ residuals align closely with physics and yield calibrated, interpretable uncertainty, whereas PEGP-LWR residuals are more orthogonal and produce nearly constant variance fields. This PEGP framework combines physical priors, uncertainty quantification, which can provide reliable support for TSE.

cs.LG↗

arXiv2025

QuantumBoost: A lazy, yet fast, quantum algorithm for learning with weak hypotheses

The technique of combining multiple votes to enhance the quality of a decision is the core of boosting algorithms in machine learning. In particular, boosting provably increases decision quality by combining multiple weak learners-hypotheses that are only slightly better than random guessing-into a single strong learner that classifies data well. There exist various versions of boosting algorithms, which we improve upon through the introduction of QuantumBoost. Inspired by classical work by Barak, Hardt and Kale, our QuantumBoost algorithm achieves the best known runtime over other boosting methods through two innovations. First, it uses a quantum algorithm to compute approximate Bregman projections faster. Second, it combines this with a lazy projection strategy, a technique from convex optimization where projections are performed infrequently rather than every iteration. To our knowledge, QuantumBoost is the first algorithm, classical or quantum, to successfully adopt a lazy projection strategy in the context of boosting.

quant-ph↗

arXiv2025

Fine-Grained Complexity via Quantum Natural Proofs

Buhrman, Patro, and Speelman presented a framework of conjectures that together form a quantum analogue of the strong exponential-time hypothesis and its variants. They called it the QSETH framework. In this paper, using a notion of quantum natural proofs (built from natural proofs introduced by Razborov and Rudich), we show how part of the QSETH conjecture that requires properties to be `compression oblivious' can in many cases be replaced by assuming the existence of quantum-secure pseudorandom functions, a standard hardness assumption. Combined with techniques from Fourier analysis of Boolean functions, we show that properties such as PARITY and MAJORITY are compression oblivious for certain circuit class $Λ$ if subexponentially secure quantum pseudorandom functions exist in $Λ$, answering an open question in [Buhrman-Patro-Speelman 2021].

quant-ph↗

arXiv2025

Improved Classical and Quantum Algorithms for the Shortest Vector Problem via Bounded Distance Decoding

The most important computational problem on lattices is the Shortest Vector Problem (SVP). In this paper, we present new algorithms that improve the state-of-the-art for provable classical/quantum algorithms for SVP. We present the following results. $\bullet$ A new algorithm for SVP that provides a smooth tradeoff between time complexity and memory requirement. For any positive integer $4\leq q\leq \sqrt{n}$, our algorithm takes $q^{13n+o(n)}$ time and requires $poly(n)\cdot q^{16n/q^2}$ memory. This tradeoff which ranges from enumeration ($q=\sqrt{n}$) to sieving ($q$ constant), is a consequence of a new time-memory tradeoff for Discrete Gaussian sampling above the smoothing parameter. $\bullet$ A quantum algorithm for SVP that runs in time $2^{0.950n+o(n)}$ and requires $2^{0.5n+o(n)}$ classical memory and poly(n) qubits. In Quantum Random Access Memory (QRAM) model this algorithm takes only $2^{0.835n+o(n)}$ time and requires a QRAM of size $2^{0.293n+o(n)}$, poly(n) qubits and $2^{0.5n}$ classical space. This improves over the previously fastest classical (which is also the fastest quantum) algorithm due to [ADRS15] that has a time and space complexity $2^{n+o(n)}$. $\bullet$ A classical algorithm for SVP that runs in time $2^{1.669n+o(n)}$ time and $2^{0.5n+o(n)}$ space. This improves over an algorithm of [CCL18] that has the same space complexity. The time complexity of our classical and quantum algorithms are obtained using a known upper bound on a quantity related to the lattice kissing number which is $2^{0.402n}$. We conjecture that for most lattices this quantity is a $2^{o(n)}$. Assuming that this is the case, our classical algorithm runs in time $2^{1.292n+o(n)}$, our quantum algorithm runs in time $2^{0.750n+o(n)}$ and our quantum algorithm in QRAM model runs in time $2^{0.667n+o(n)}$.

cs.DS↗

arXiv2025

QSETH strikes again: finer quantum lower bounds for lattice problem, strong simulation, hitting set problem, and more

While seemingly undesirable, it is not a surprising fact that there are certain problems for which quantum computers offer no computational advantage over their respective classical counterparts. Moreover, there are problems for which there is no `useful' computational advantage possible with the current quantum hardware. This situation however can be beneficial if we don't want quantum computers to solve certain problems fast - say problems relevant to post-quantum cryptography. In such a situation, we would like to have evidence that it is difficult to solve those problems on quantum computers; but what is their exact complexity? To do so one has to prove lower bounds, but proving unconditional time lower bounds has never been easy. As a result, resorting to conditional lower bounds has been quite popular in the classical community and is gaining momentum in the quantum community. In this paper, by the use of the QSETH framework [Buhrman-Patro-Speelman 2021], we are able to understand the quantum complexity of a few natural variants of CNFSAT, such as parity-CNFSAT or counting-CNFSAT, and also are able to comment on the non-trivial complexity of approximate-#CNFSAT; both of these have interesting implications about the complexity of (variations of) lattice problems, strong simulation and hitting set problem, and more. In the process, we explore the QSETH framework in greater detail than was (required and) discussed in the original paper, thus also serving as a useful guide on how to effectively use the QSETH framework.

quant-ph↗

arXiv2025

Superior Frequency Stability and Long-Lived State-Swapping in Cubic-SiC Mechanical Mode Pairs

The multimode cavity optomechanical system offers versatile applications including state transduction, coherent interconnection, and many-body simulations. In this study, we developed a cavity electromechanical system that integrates a 3C-SiC membrane and a rectangular superconducting cavity to observe the generation of nearly degenerate pairs of mechanical modes. Subsequently, we derive the expression for intrinsic frequency under nonuniform stress and find that this method supports a remarkably resolution for stress analysis in thin films. Experimentally, we perform collective fitting on the measured set of 57 mechanical modes, revealing deviations in biaxial non-uniform stress on the order of MPa. These degeneracy-broken mechanical modes exhibit exceptional quality factors as high as $10^8$ in a thermal bath of 10 mK. Furthermore, Allan deviation indicates that these modes exhibit extremely stable frequencies compared with different types of optomechanical devices. We then performed state-swapping between near-degenerate mode pairs, demonstrating the transfer efficiency exceeding 78\%, attributed to their exceptionally long lifetimes. This study paves the way for the design of compact quantum phononic devices featuring high-quality-factor mechanical multimode.

quant-ph↗

arXiv2025

MReg: A Novel Regression Model with MoE-based Video Feature Mining for Mitral Regurgitation Diagnosis

Color Doppler echocardiography is a crucial tool for diagnosing mitral regurgitation (MR). Recent studies have explored intelligent methods for MR diagnosis to minimize user dependence and improve accuracy. However, these approaches often fail to align with clinical workflow and may lead to suboptimal accuracy and interpretability. In this study, we introduce an automated MR diagnosis model (MReg) developed on the 4-chamber cardiac color Doppler echocardiography video (A4C-CDV). It follows comprehensive feature mining strategies to detect MR and assess its severity, considering clinical realities. Our contribution is threefold. First, we formulate the MR diagnosis as a regression task to capture the continuity and ordinal relationships between categories. Second, we design a feature selection and amplification mechanism to imitate the sonographer's diagnostic logic for accurate MR grading. Third, inspired by the Mixture-of-Experts concept, we introduce a feature summary module to extract the category-level features, enhancing the representational capacity for more accurate grading. We trained and evaluated our proposed MReg on a large in-house A4C-CDV dataset comprising 1868 cases with three graded regurgitation labels. Compared to other weakly supervised video anomaly detection and supervised classification methods, MReg demonstrated superior performance in MR diagnosis. Our code is available at: https://github.com/cskdstz/MReg.

cs.CV↗

arXiv2025

Research on evolution and early warning model of network public opinion based on online Latent Dirichlet distribution model and BP neural network

Online public opinion is increasingly becoming a significant factor affecting the stability of the internet and society, particularly as the frequency of online public opinion crises has risen in recent years. Enhancing the capability for early warning of online public opinion crises is urgent. The most effective approach is to identify potential crises in their early stages and implement corresponding management measures. This study establishes a preliminary indicator system for online public opinion early warning, based on the principles of indicator system construction and the characteristics and evolution patterns of online public opinion. Subsequently, data-driven methodologies were employed to collect and preprocess public opinion indicator data. Utilizing grey relational analysis and the K-Means clustering algorithm, we classified online public opinion events into three levels: slight, warning, and severe. Furthermore, we constructed an online topic evolution model using the online Hierarchical Dirichlet Process model to analyze the thematic changes of online public opinion events across different warning levels. Finally, we developed an online public opinion early warning model using a Backpropagation (BP) neural network. The test results of early warning samples show that the model achieves high accuracy. Thus, in practical early warning applications, the BP neural network can be effectively utilized for predicting online public opinion events.

cs.SI↗

arXiv2025

Learning to Hop for a Single-Legged Robot with Parallel Mechanism

This work presents the application of reinforcement learning to improve the performance of a highly dynamic hopping system with a parallel mechanism. Unlike serial mechanisms, parallel mechanisms can not be accurately simulated due to the complexity of their kinematic constraints and closed-loop structures. Besides, learning to hop suffers from prolonged aerial phase and the sparse nature of the rewards. To address them, we propose a learning framework to encode long-history feedback to account for the under-actuation brought by the prolonged aerial phase. In the proposed framework, we also introduce a simplified serial configuration for the parallel design to avoid directly simulating parallel structure during the training. A torque-level conversion is designed to deal with the parallel-serial conversion to handle the sim-to-real issue. Simulation and hardware experiments have been conducted to validate this framework.

cs.RO↗

arXiv2024

A Quantum Speed-Up for Approximating the Top Eigenvectors of a Matrix

Finding a good approximation of the top eigenvector of a given $d\times d$ matrix $A$ is a basic and important computational problem, with many applications. We give two different quantum algorithms that, given query access to the entries of a Hermitian matrix $A$ and assuming a constant eigenvalue gap, output a classical description of a good approximation of the top eigenvector: one algorithm with time complexity $\mathcal{\tilde{O}}(d^{1.75})$ and one with time complexity $d^{1.5+o(1)}$ (the first algorithm has a slightly better dependence on the $\ell_2$-error of the approximating vector than the second, and uses different techniques of independent interest). Both of our quantum algorithms provide a polynomial speed-up over the best-possible classical algorithm, which needs $Ω(d^2)$ queries to entries of $A$, and hence $Ω(d^2)$ time. We extend this to a quantum algorithm that outputs a classical description of the subspace spanned by the top-$q$ eigenvectors in time $qd^{1.5+o(1)}$. We also prove a nearly-optimal lower bound of $\tildeΩ(d^{1.5})$ on the quantum query complexity of approximating the top eigenvector. Our quantum algorithms run a version of the classical power method that is robust to certain benign kinds of errors, where we implement each matrix-vector multiplication with small and well-behaved error on a quantum computer, in different ways for the two algorithms. Our first algorithm estimates the matrix-vector product one entry at a time, using a new "Gaussian phase estimation" procedure. Our second algorithm uses block-encoding techniques to compute the matrix-vector product as a quantum state, from which we obtain a classical description by a new time-efficient unbiased pure-state tomography procedure.

quant-ph↗

arXiv2024

Optimizing Transformer based on high-performance optimizer for predicting employment sentiment in American social media content

This article improves the Transformer model based on swarm intelligence optimization algorithm, aiming to predict the emotions of employment related text content on American social media. Through text preprocessing, feature extraction, and vectorization, the text data was successfully converted into numerical data and imported into the model for training. The experimental results show that during the training process, the accuracy of the model gradually increased from 49.27% to 82.83%, while the loss value decreased from 0.67 to 0.35, indicating a significant improvement in the performance of the model on the training set. According to the confusion matrix analysis of the training set, the accuracy of the training set is 86.15%. The confusion matrix of the test set also showed good performance, with an accuracy of 82.91%. The accuracy difference between the training set and the test set is only 3.24%, indicating that the model has strong generalization ability. In addition, the evaluation of polygon results shows that the model performs well in classification accuracy, sensitivity, specificity, and area under the curve (AUC), with a Kappa coefficient of 0.66 and an F-measure of 0.80, further verifying the effectiveness of the model in social media sentiment analysis. The improved model proposed in this article not only improves the accuracy of sentiment recognition in employment related texts on social media, but also has important practical significance. This social media based data analysis method can not only capture social dynamics in a timely manner, but also promote decision-makers to pay attention to public concerns and provide data support for improving employment conditions.

cs.CL↗

arXiv2024

Improved Unet model for brain tumor image segmentation based on ASPP-coordinate attention mechanism

In this paper, we propose an improved Unet model for brain tumor image segmentation, which combines coordinate attention mechanism and ASPP module to improve the segmentation effect. After the data set is divided, we do the necessary preprocessing to the image and use the improved model to experiment. First, we trained and validated the traditional Unet model. By analyzing the loss curve of the training set and the validation set, we can see that the loss value continues to decline at the first epoch and becomes stable at the eighth epoch. This process shows that the model constantly optimizes its parameters to improve performance. At the same time, the change in the miou (mean Intersection over Union) index shows that the miou value exceeded 0.6 at the 15th epoch, remained above 0.6 thereafter, and reached above 0.7 at the 46th epoch. These results indicate that the basic Unet model is effective in brain tumor image segmentation. Next, we introduce an improved Unet algorithm based on coordinate attention mechanism and ASPP module for experiments. By observing the loss change curves of the training set and the verification set, it is found that the loss value reaches the lowest point at the sixth epoch and then remains relatively stable. At the same time, the miou indicator has stabilized above 0.7 since the 20th epoch and has reached a maximum of 0.76. These results show that the new mechanism introduced significantly improves the segmentation ability of the model. Finally, we apply the trained traditional Unet model and the improved Unet model based on the coordinate attention mechanism and ASPP module to the test set for brain tumor image segmentation prediction. Compared to the traditional Unet, the enhanced model offers superior segmentation and edge accuracy, providing a more reliable method for medical image analysis with the coordinate attention mechanism and ASPP module.

eess.IV↗

arXiv2024

Explainable and Controllable Motion Curve Guided Cardiac Ultrasound Video Generation

Echocardiography video is a primary modality for diagnosing heart diseases, but the limited data poses challenges for both clinical teaching and machine learning training. Recently, video generative models have emerged as a promising strategy to alleviate this issue. However, previous methods often relied on holistic conditions during generation, hindering the flexible movement control over specific cardiac structures. In this context, we propose an explainable and controllable method for echocardiography video generation, taking an initial frame and a motion curve as guidance. Our contributions are three-fold. First, we extract motion information from each heart substructure to construct motion curves, enabling the diffusion model to synthesize customized echocardiography videos by modifying these curves. Second, we propose the structure-to-motion alignment module, which can map semantic features onto motion curves across cardiac structures. Third, The position-aware attention mechanism is designed to enhance video consistency utilizing Gaussian masks with structural position information. Extensive experiments on three echocardiography datasets show that our method outperforms others regarding fidelity and consistency. The full code will be released at https://github.com/mlmi-2024-72/ECM.

eess.IV↗