arXiv subjects

Ke Lin

Publications and source records attributed to Ke Lin.

At least 19 recordsLinked to original sources

arXiv2026

Triplet-Block Diffusion RWKV

Causal Transformer language models suffer from strictly sequential decoding and a quadratic per-step attention cost. While linear-time causal models and discrete diffusion models each address these weaknesses, their integration remains inherently inconsistent: diffusion requires bidirectional attention, while causal models are unidirectional. To unify these architectures, we propose $B^3D-RWKV$, a diffusion RWKV variant that integrates the model's $O(L)$ inference efficiency with parallel, bidirectional discrete-diffusion through a \emph{triplet-block layout} method. $B^3D-RWKV-7.2B$ reaches comparable accuracy on an 8-task suite versus existing models while significantly outperforming baselines in decoding throughput with an average of $\mathbf{1.6\times}$ speedup.

cs.CL↗

arXiv2026

Interband Berry connection measurement in the optical honeycomb lattice

The geometry of Bloch bands affects many physical properties of crystalline solids and other spatially periodic systems. Direct experimental determination of such geometry is an active area of research. In this work, we focus on the fundamental connection between optical excitations and the relative geometry of pairs of Bloch bands, as characterized by the interband Berry connection. We simulate the response of electrons in solids to optical excitation by the response of ultracold fermionic atoms in optical lattices to periodic modulation of the lattice position. The strength of resonant excitation between bands, measured at each quasimomentum and for various lattice-shaking polarizations, directly maps out the interband Berry connection. We apply this method to the optical honeycomb lattice, driving excitations between the ground $n=1$ band and the excited $n'=\{2,3,4\}$ bands. We observe transparency lines of quasimomenta at which the response to excitation of specific polarization is zero. Further, the interband Berry connection between bands 1 and 3 shows irreducible Dirac strings connecting the $K$ and $K'$ points in the Brillouin zone, lines along which the interband Berry connection abruptly changes orientation. Our work establishes optical response as a powerful tool for characterizing geometrical and topological properties of band structure.

cond-mat.quant-gas↗

arXiv2026

SciDER: Scientific Data-centric End-to-end Researcher

While large language models accelerate scientific discovery, existing agents face severe limitations in adaptability, domain generalization, and multimodal scalability, often struggling to autonomously process raw, domain-specific experimental data. To overcome these barriers, we introduce SciDER, a multi-agent system designed to flexibly automate the entire research lifecycle. This framework employs a novel data-centric approach and integrates a dynamic multimodal skill system across four specialized sub-agents. Specifically, an ideation agent generates novel hypotheses via Evolutionary Idea Search, a data analysis agent systematically structures raw data, an experimentation agent synthesizes executable code grounded in dataset characteristics, and a critic agent drives iterative self-refinement. To democratize open-source scientific discovery, we release OpenSciDER-SFT-8K, a high-quality execution trajectory dataset, alongside the OpenSciDER-27B fine-tuned model. Across six benchmarks, SciDER and OpenSciDER obtain competitive or leading results, with especially strong gains on data-centric analysis, end-to-end research execution, and multimodal scientific visualization. By integrating data analysis with experimental execution, SciDER bridges the gap between abstract scientific reasoning and reproducible experimentation synthesis.

cs.AI↗

arXiv2025

Wavelet-based Disentangled Adaptive Normalization for Non-stationary Times Series Forecasting

Forecasting non-stationary time series is a challenging task because their statistical properties often change over time, making it hard for deep models to generalize well. Instance-level normalization techniques can help address shifts in temporal distribution. However, most existing methods overlook the multi-component nature of time series, where different components exhibit distinct non-stationary behaviors. In this paper, we propose Wavelet-based Disentangled Adaptive Normalization (WDAN), a model-agnostic framework designed to address non-stationarity in time series forecasting. WDAN uses discrete wavelet transforms to break down the input into low-frequency trends and high-frequency fluctuations. It then applies tailored normalization strategies to each part. For trend components that exhibit strong non-stationarity, we apply first-order differencing to extract stable features used for predicting normalization parameters. Extensive experiments on multiple benchmarks demonstrate that WDAN consistently improves forecasting accuracy across various backbone model. Code is available at this repository: https://github.com/MonBG/WDAN.

cs.LG↗

arXiv2025

Think When You Need: Self-Adaptive Chain-of-Thought Learning

Chain of Thought (CoT) reasoning enhances language models' performance but often leads to inefficient "overthinking" on simple problems. We identify that existing approaches directly penalizing reasoning length fail to account for varying problem complexity. Our approach constructs rewards through length and quality comparisons, guided by theoretical assumptions that jointly enhance solution correctness with conciseness. Moreover, we further demonstrate our method to fuzzy tasks where ground truth is unavailable. Experiments across multiple reasoning benchmarks demonstrate that our method maintains accuracy while generating significantly more concise explanations, effectively teaching models to "think when needed."

cs.CL↗

arXiv2024

Towards Precise Scaling Laws for Video Diffusion Transformers

Achieving optimal performance of video diffusion transformers within given data and compute budget is crucial due to their high training costs. This necessitates precisely determining the optimal model size and training hyperparameters before large-scale training. While scaling laws are employed in language models to predict performance, their existence and accurate derivation in visual generation models remain underexplored. In this paper, we systematically analyze scaling laws for video diffusion transformers and confirm their presence. Moreover, we discover that, unlike language models, video diffusion models are more sensitive to learning rate and batch size, two hyperparameters often not precisely modeled. To address this, we propose a new scaling law that predicts optimal hyperparameters for any model size and compute budget. Under these optimal settings, we achieve comparable performance and reduce inference costs by 40.1% compared to conventional scaling methods, within a compute budget of 1e10 TFlops. Furthermore, we establish a more generalized and precise relationship among validation loss, any model size, and compute budget. This enables performance prediction for non-optimal model sizes, which may also be appealed under practical inference cost constraints, achieving a better trade-off.

cs.CV↗

arXiv2024

Koala-36M: A Large-scale Video Dataset Improving Consistency between Fine-grained Conditions and Video Content

With the continuous progress of visual generation technologies, the scale of video datasets has grown exponentially. The quality of these datasets plays a pivotal role in the performance of video generation models. We assert that temporal splitting, detailed captions, and video quality filtering are three crucial determinants of dataset quality. However, existing datasets exhibit various limitations in these areas. To address these challenges, we introduce Koala-36M, a large-scale, high-quality video dataset featuring accurate temporal splitting, detailed captions, and superior video quality. The essence of our approach lies in improving the consistency between fine-grained conditions and video content. Specifically, we employ a linear classifier on probability distributions to enhance the accuracy of transition detection, ensuring better temporal consistency. We then provide structured captions for the splitted videos, with an average length of 200 words, to improve text-video alignment. Additionally, we develop a Video Training Suitability Score (VTSS) that integrates multiple sub-metrics, allowing us to filter high-quality videos from the original corpus. Finally, we incorporate several metrics into the training process of the generation model, further refining the fine-grained conditions. Our experiments demonstrate the effectiveness of our data processing pipeline and the quality of the proposed Koala-36M dataset. Our dataset and code have been released at https://koala36m.github.io/.

cs.CV↗

arXiv2024

ViRED: Prediction of Visual Relations in Engineering Drawings

To accurately understand engineering drawings, it is essential to establish the correspondence between images and their description tables within the drawings. Existing document understanding methods predominantly focus on text as the main modality, which is not suitable for documents containing substantial image information. In the field of visual relation detection, the structure of the task inherently limits its capacity to assess relationships among all entity pairs in the drawings. To address this issue, we propose a vision-based relation detection model, named ViRED, to identify the associations between tables and circuits in electrical engineering drawings. Our model mainly consists of three parts: a vision encoder, an object encoder, and a relation decoder. We implement ViRED using PyTorch to evaluate its performance. To validate the efficacy of ViRED, we conduct a series of experiments. The experimental results indicate that, within the engineering drawing dataset, our approach attained an accuracy of 96\% in the task of relation prediction, marking a substantial improvement over existing methodologies. The results also show that ViRED can inference at a fast speed even when there are numerous objects in a single engineering drawing.

cs.CV↗

arXiv2024

Sharp critical mass criteria for weak solutions to a degenerate cross-attraction system

The qualitative study of solutions to the coupled parabolic-elliptic chemotaxis system with nonlinear diffusion for two species will be considered in the whole Euclidean space $\mathbb{R}^d$ ($d\geq 3$). It was proven in \cite{CK2021-ANA} that there exist two critical curves that separate the global existence and blow-up of weak solutions to the above problem. We improve this result by providing sharp criteria for the dicothomy: global existence of weak solution versus blow-up below and at these curves. Besides, there exist sharp critical masses of initial data at the intersection of the two critical lines, which extend the well-known critical mass phenomenon in one-species Keller-Segel system in \cite{BCL09-CVPDE} to two-species case.

math.AP↗

arXiv2024

SEA: Supervised Embedding Alignment for Token-Level Visual-Textual Integration in MLLMs

Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities by integrating visual and textual inputs, yet modality alignment remains one of the most challenging aspects. Current MLLMs typically rely on simple adapter architectures and pretraining approaches to bridge vision encoders with large language models (LLM), guided by image-level supervision. We identify this paradigm often leads to suboptimal alignment between modalities, significantly constraining the LLM's ability to properly interpret and reason with visual features particularly for smaller language models. This limitation degrades overall performance-particularly for smaller language models where capacity constraints are more pronounced and adaptation capabilities are limited. To address this fundamental limitation, we propose Supervised Embedding Alignment (SEA), a token-level supervision alignment method that enables more precise visual-text alignment during pretraining. SEA introduces minimal computational overhead while preserving language capabilities and substantially improving cross-modal understanding. Our comprehensive analyses reveal critical insights into the adapter's role in multimodal integration, and extensive experiments demonstrate that SEA consistently improves performance across various model sizes, with smaller models benefiting the most (average performance gain of 7.61% for Gemma-2B). This work establishes a foundation for developing more effective alignment strategies for future multimodal systems.

cs.CV↗

arXiv2024

Low-Latency Privacy-Preserving Deep Learning Design via Secure MPC

Secure multi-party computation (MPC) facilitates privacy-preserving computation between multiple parties without leaking private information. While most secure deep learning techniques utilize MPC operations to achieve feasible privacy-preserving machine learning on downstream tasks, the overhead of the computation and communication still hampers their practical application. This work proposes a low-latency secret-sharing-based MPC design that reduces unnecessary communication rounds during the execution of MPC protocols. We also present a method for improving the computation of commonly used nonlinear functions in deep learning by integrating multivariate multiplication and coalescing different packets into one to maximize network utilization. Our experimental results indicate that our method is effective in a variety of settings, with a speedup in communication latency of $10\sim20\%$.

cs.CR↗

arXiv2024

Chaos-Assisted Dynamical Tunneling in Flat Band Superwires

Recent theoretical investigations have revealed unconventional transport mechanisms within high Brilliouin zones of two-dimensional superlattices. Electrons can navigate along channels we call superwires, gently guided without brute force confinement. Such dynamical confinement is caused by weak superlattice deflections, markedly different from the static or energetic confinement observed in traditional wave guides or one-dimensional electron wires. The quantum properties of superwires give rise to elastic dynamical tunneling, linking disjoint regions of the corresponding classical phase space, and enabling the emergence of several parallel channels. This paper provides the underlying theory and mechanisms that facilitate dynamical tunneling assisted by chaos in periodic lattices. Moreover, we show that the mechanism of dynamical tunneling can be effectively conceptualized through the lens of a paraxial approximation. Our results further reveal that superwires predominantly exist within flat bands, emerging from eigenstates that represent linear combinations of conventional degenerate Bloch states. Finally, we quantify tunneling rates across various lattice configurations, and demonstrate the tunneling can be suppressed in a controlled fashion, illustrating potential implications in future nanodevices.

cond-mat.mes-hall↗

arXiv2024

Zero-shot Generative Linguistic Steganography

Generative linguistic steganography attempts to hide secret messages into covertext. Previous studies have generally focused on the statistical differences between the covertext and stegotext, however, ill-formed stegotext can readily be identified by humans. In this paper, we propose a novel zero-shot approach based on in-context learning for linguistic steganography to achieve better perceptual and statistical imperceptibility. We also design several new metrics and reproducible language evaluations to measure the imperceptibility of the stegotext. Our experimental results indicate that our method produces $1.926\times$ more innocent and intelligible stegotext than any other method.

cs.CL↗

arXiv2024

Lost in Overlap: Exploring Logit-based Watermark Collision in LLMs

The proliferation of large language models (LLMs) in generating content raises concerns about text copyright. Watermarking methods, particularly logit-based approaches, embed imperceptible identifiers into text to address these challenges. However, the widespread usage of watermarking across diverse LLMs has led to an inevitable issue known as watermark collision during common tasks, such as paraphrasing or translation. In this paper, we introduce watermark collision as a novel and general philosophy for watermark attacks, aimed at enhancing attack performance on top of any other attacking methods. We also provide a comprehensive demonstration that watermark collision poses a threat to all logit-based watermark algorithms, impacting not only specific attack scenarios but also downstream applications.

cs.CL↗

arXiv2023

Skipping Scheme for Gate-hiding Garbled Circuits

In classic settings of garbled circuits, each gate type is leaked to improve both space and speed optimization. Zahur et al. have shown in EUROCRYPT 2015 that a typical linear garbling scheme requires at least two $\lambda$-bit elements per gate with a security parameter of $\lambda$, which limits their efficiency. In contrast to typical garbled circuits, gate-hiding garbled circuits have the potential to drastically reduce time costs, although they have been underappreciated. We propose the first skipping scheme for gate-hiding garbled circuits to enhance the efficiency of evaluation by observing prime implicants. Our scheme introduces skip gates to eliminate the need to calculate the entire circuit, enabling unnecessary execution paths to be avoided. We also introduce two variants of our scheme that balance security with parallelism. A proof of hybrid security that combines simulation-based and symmetry-based security in semi-honest scenarios is presented to demonstrate its security under gate-hiding conditions. Our scheme will inspire new directions to improve the general garbling scheme and lead to more practical ones.

cs.CR↗

arXiv2023

Context-Aware Indoor Point Cloud Object Generation through User Instructions

Indoor scene modification has emerged as a prominent area within computer vision, particularly for its applications in Augmented Reality (AR) and Virtual Reality (VR). Traditional methods often rely on pre-existing object databases and predetermined object positions, limiting their flexibility and adaptability to new scenarios. In response to this challenge, we present a novel end-to-end multi-modal deep neural network capable of generating point cloud objects seamlessly integrated with their surroundings, driven by textual instructions. Our model revolutionizes scene modification by enabling the creation of new environments with previously unseen object layouts, eliminating the need for pre-stored CAD models. Leveraging Point-E as our generative model, we introduce innovative techniques such as quantized position prediction and Top-K estimation to address the issue of false negatives resulting from ambiguous language descriptions. Furthermore, we conduct comprehensive evaluations to showcase the diversity of generated objects, the efficacy of textual instructions, and the quantitative metrics, affirming the realism and versatility of our model in generating indoor objects. To provide a holistic assessment, we incorporate visual grounding as an additional metric, ensuring the quality and coherence of the scenes produced by our model. Through these advancements, our approach not only advances the state-of-the-art in indoor scene modification but also lays the foundation for future innovations in immersive computing and digital environment creation.

cs.CV↗

arXiv2022

WPNAS: Neural Architecture Search by jointly using Weight Sharing and Predictor

Weight sharing based and predictor based methods are two major types of fast neural architecture search methods. In this paper, we propose to jointly use weight sharing and predictor in a unified framework. First, we construct a SuperNet in a weight-sharing way and probabilisticly sample architectures from the SuperNet. To increase the correctness of the evaluation of architectures, besides direct evaluation using the inherited weights, we further apply a few-shot predictor to assess the architecture on the other hand. The final evaluation of the architecture is the combination of direct evaluation, the prediction from the predictor and the cost of the architecture. We regard the evaluation as a reward and apply a self-critical policy gradient approach to update the architecture probabilities. To further reduce the side effects of weight sharing, we propose a weakly weight sharing method by introducing another HyperNet. We conduct experiments on datasets including CIFAR-10, CIFAR-100 and ImageNet under NATS-Bench, DARTS and MobileNet search space. The proposed WPNAS method achieves state-of-the-art performance on these datasets.

cs.LG↗

arXiv2021

Classical-to-quantum transition in multimode nonlinear systems with strong photon-photon coupling

With advanced micro- and nano-photonic structures, the vacuum photon-photon coupling rate is anticipated to approach the intrinsic loss rate and lead to unconventional quantum effects. Here, we investigate the classical-to-quantum transition of such photonic nonlinear systems using the quantum cluster-expansion method, which addresses the computational challenge in tracking large photon number states of the fundamental and harmonic optical fields involved in the second harmonic generation process. Compared to the mean-field approximation used in weak coupling limit, the quantum cluster-expansion method solves multimode dynamics efficiently and reveals the quantum behaviors of optical parametric oscillations around the threshold. This work presents a universal tool to study quantum dynamics of multimode systems and explore the nonlinear photonic devices for continuous-variable quantum information processing.

quant-ph↗