Searcharxiv⌕ Search

arXiv subjects

Jun Yu

Publications and source records attributed to Jun Yu.

At least 37 records · Page 2Linked to original sources
arXiv2026

Hi-Singers: A Comprehensive High-Quality Dataset for Expressive Audio-Driven Singing Head Synthesis

State-of-the-art models for audio-driven digital human generation have achieved photo-realistic results in talking-head synthesis. However, extending these models to singing-head synthesis remains challenging due to a significant Domain Gap: singing requires more exaggerated expressions, vivid jaw openings, and precise rhythmic synchronization. Current models, primarily trained on speech datasets, often struggle with "rhythmic drift" and constrained dynamics. To address this, we introduce Hi-Singers, the first large-scale, high-quality, in-the-wild video dataset specifically tailored for singing head synthesis. Hi-Singers undergoes a rigorous automated and manual filtering pipeline, ensuring strict thematic adherence, high-resolution rendering, and stable motion, resulting in 29,608 video segments totaling approximately 170 hours. We further establish a dedicated evaluation benchmark balanced across linguistic and musical styles. Extensive experiments across diverse architectures, including 3D-coefficient and diffusion-based models, demonstrate that Hi-Singers consistently and significantly improves performance across all dimensions. Specifically, it enables models to achieve superior visual realism, enhanced lip-sync consistency, and more precise rhythmic dynamics, effectively bridging the domain gap and setting a new performance standard for the singing synthesis task. The dataset is available at https://huggingface.co/datasets/CharlesZhang-USTC/Hi-Singers

cs.MM↗
arXiv2026

KaLM-Embedding-V2: Superior Training Techniques and Data Inspire A Versatile Embedding Model

Recent advancements in Large Language Models (LLMs)-based text embedding models primarily focus on data scaling or synthesis, yet limited exploration of training techniques and data quality, thereby constraining performance. In this work, we propose KaLM-Embedding-V2 from the Lychee-KaLM team, a series of versatile and compact embedding models, systematically incentivizing advanced embedding capability in LLMs by superior training techniques and high-quality data. For model architecture, we implement the models in a 0.5B compact size with simple mean-pooling to produce fixed-length embeddings and remove the causal attention mask to enable fully bidirectional representation learning. For training techniques, we propose a progressive multi-stage training pipeline: pre-training on weakly supervised large-scale datasets, fine-tuning with supervised high-quality datasets, and contrastive distillation with fine-grained soft signals, integrated with focal-style reweighting and online hard-negative mixing to emphasize difficult samples and enrich hard negatives, respectively. For training data, we curate over 20 categories for pre-training and 100 categories for fine-tuning and contrastive distillation to improve both performance and generalization, leveraging task-specific instructions, hard-negative mining, and example-based multi-class labeling to ensure high quality. Combining these techniques, our KaLM-Embedding-V2 series achieves state-of-the-art performance on the Massive Text Embedding Benchmark, outperforming models of comparable size and rivaling models 3--26x larger, setting a new standard for versatile and compact embedding models under 1B parameters. The code, data, and models are available at https://kalm-embedding.github.io/.

cs.CL↗
arXiv2026

A Token is Worth over 1,000 Tokens: Efficient Knowledge Distillation through Low-Rank Clone

Training high-performing Small Language Models (SLMs) remains costly, even with knowledge distillation and pruning from larger teacher models. Existing work often faces three key challenges: (1) information loss from hard pruning, (2) inefficient alignment of representations, and (3) underutilization of informative activations, particularly from Feed-Forward Networks (FFNs). To address these challenges, we introduce Low-Rank Clone (LRC), an efficient pre-training method that constructs SLMs aspiring to behavioral equivalence with strong teacher models. LRC trains a set of low-rank projection matrices that jointly enable soft pruning by compressing teacher weights, and activation clone by aligning student activations, including FFN signals, with those of the teacher. This unified design maximizes knowledge transfer while removing the need for explicit alignment modules. Extensive experiments with open-source teachers (e.g., Llama-3.2-3B-Instruct, Qwen2.5-3B/7B-Instruct) show that LRC matches or surpasses state-of-the-art models trained on trillions of tokens--while using only 20B tokens, achieving over 1,000x training efficiency. Our codes and model checkpoints are available at https://github.com/CURRENTF/LowRankClone and https://huggingface.co/collections/JitaiHao/low-rank-clone-lrc-6828389e96a93f1d4219dfaf.

cs.CL↗
arXiv2026

Metis: Typed Runtime Mediation for Tool-Using Software Agents

Software agents connect probabilistic model output to operations that change repositories, processes, networks, and graphical applications. We present Metis, a multi-provider runtime that converts provider streams into typed events before admitted calls reach external effects. Its execution path makes permission decisions, interference classes, terminal results, and lifecycle transitions explicit and inspectable. We evaluate these mechanisms on frozen source artifacts. Across 30 matched real-I/O pairs, four-class mediation reduced median elapsed time from 25.958 ms under forced serialization to 14.146 ms. The mean paired difference was -12.295 ms (95% bootstrap interval [-12.968, -11.694]), with mediation faster in all pairs. A ten-case fault matrix exposed duplicate-identifier and rollback limits. In a child-boundary ablation, the full gate-plus-registry condition blocked the declared unauthorized effect and hid all five escape tools. Removing both protections reversed both observations. A decision-only permission oracle matched all ten declared cases across five invocation routes. Five model conditions also completed a fixed Read-marker protocol in 3/3 trials each. These results support bounded claims about dispatch, permission routing, child authority, and provider-valid trace closure. They do not establish model competence, semantic safety, rollback, or superiority over another runtime.

cs.SE↗
arXiv2026

STA-VPT: SpatioTemporally Aligned Visual Prompt Tuning

Typical methods for visual prompt tuning follow the sequential modeling paradigm originating from NLP, learning a sequence of unordered parameterized tokens as visual prompts, which are then prefixed to the flattened image representation for model adaptation. While such a sequential prompting paradigm has exhibited great promise, it presents two potential limitations. First, the learned visual prompts, presented in an unordered sequential form, are unable to capture the underlying spatial relations in the input image which are crucial for effective image encoding. Second, all prompt tokens serve the same role by performing uniform prompting for all image tokens without distinction, lacking fine-grained prompting capability\---i.e., individualized prompting for different visual tokens to capture region-specific semantic patterns. In this work, we introduce the \mymodel model (\emph{STA-VPT}), a novel visual prompting paradigm, which learns a two-dimensional prompt token map for image prompting or a three-dimensional token volume for video prompting, ensuring spatial (or spatiotemporal) alignment with the input image token map (or video token volume). This alignment enables the visual prompts to preserve the spatial (or spatiotemporal) structure, thereby learning the underlying relations within the visual input. Furthermore, each prompt token serves as a specialized prompting expert and is designated to exclusively prompt for the spatially (or spatiotemporally) corresponding visual tokens. Consequently, our \emph{STA-VPT} is capable of performing individualized prompting for different spatiotemporal regions, potentially improving the prompting performance through fine-grained allocation of prompting capacity, in line with mixture of experts (MoE) principles.

cs.CV↗
arXiv2026

VersaDB: A High-Performance AI Storage Database for Unifying Mutimodal Datasets

The AI field has been rapidly developing, leading to the emergence of a large number of AI training datasets of various types. These datasets contain different modalities, including text, images, audio, etc., and may come in various data storage formats. With the advancement of AI hardware, AI computation units like GPUs, TPUs, and NPUs can greatly accelerate the training speed of AI models, which in turn increases the demand for faster data processing. When using existing AI processing frameworks to handle datasets with different modalities and storage formats, processing speeds may be suboptimal due to issues such as data layout and the way users handle the data. Therefore, using a unified database to store multiple data formats can better manage and optimize data access. In this paper, we introduce VersaDB, a database designed specifically for AI datasets with various modalities. We implemented a page-based storage system, separating structured and unstructured data. Additionally, we generated B+ tree-based index files to accelerate data access. VersaDB supports automatic sharding and maintains a hierarchical metadata management system, with corresponding metadata maintained at the page, shard, and global levels, forming the foundation for the efficient operation of the database. We also focused on ease of use by providing APIs for directly converting datasets into VersaDB, as well as APIs for converting popular AI data storage formats (e.g., CSV, TFRecord, .bin) into VersaDB.Our experiments show that using VersaDB can achieve up to 5.35x acceleration and maintain consistent performance across different parallelism levels.

cs.CV↗
arXiv2026

Trustworthy Decisions in Reliability Set Estimation under Insufficient Model Information

Reliability set estimation identifies input regions where a response probability exceeds a target level, bridging estimation and safety-critical decisions. Practitioners typically start with a working model, an imperfect approximation of the true response surface. Relying on this imperfect model may incur decision risk, potentially certifying unsafe regions as safe. We develop a unified framework that turns such a working model into a trustworthy decision rule. First, a modeling-then-calibration procedure decouples estimation from decision. Since the true set is unobservable, we introduce an asymmetric, observable surrogate loss and use a separate calibration set to select a bias-correcting threshold, reducing decision risk and achieving $O_P(1/n)$ volume convergence. Second, we leverage conformal risk control with the surrogate loss to control false inclusion risk, which is the most safety-critical error, at a pre-specified level regardless of working model quality. Together, these calibration procedures show that a separate calibration set is necessary for risk control. Third, an adaptive design concentrates observations on the reliability set and its boundary, improving model quality where errors most affect decisions while controlling budget elsewhere. Numerical studies show not only more accurate set estimates but also calibrated finite-sample risk control that classical plug-in methods lack.

math.ST↗
arXiv2026

Modeling and Forecasting Realized Volatility with Multivariate Fractional Brownian Motion

A multivariate fractional Brownian motion (mfBm) with component-wise Hurst exponents is used to model and forecast realized volatility (RV). We investigate the interplay between correlation coefficients and Hurst exponents and propose a novel method to estimate model parameters, establishing its consistency and asymptotic normality. Additionally, we develop a time-reversibility test, which is typically not rejected by RV data. When the data generating process is a time-reversible mfBm, we derive optimal forecasting formulae and analyze their properties. A key insight is that an mfBm with different Hurst exponents and non-zero correlations can reduce forecasting errors compared to a one-dimensional model. Consistent with this theory, out-of-sample forecasts using the time-reversible mfBm show improvements over univariate fBm, particularly when the estimated Hurst exponents differ significantly. Empirical results demonstrate that mfBm outperforms HAR and its variants in terms of out-of-sample forecast.

q-fin.ST↗
arXiv2026

WBMM: Windowed Batch Matrix Multiplication for Efficient Large Receptive Field Convolution

Large kernel depthwise convolutions achieve strong performance but suffer from significant degradation as kernel size grows due to irregular memory access from gather-based computation; while Large Kernel Acceleration (LKA) helps on small feature maps, it becomes counterproductive on large feature maps, even slower than non-accelerated implementations. We propose Windowed Batch Matrix Multiplication (WBMM), which partitions input into contiguous windows and indexes a compact relative position bias table to construct weight matrices, enabling regular memory access via batched matrix multiplication. This yields a unique property: WBMM's throughput improves with larger windows, opposite to depthwise convolutions that degrade with larger kernels. Operator-level benchmarks show WBMM with 14x14 windows outperforms 5x5 depthwise convolution baselines in speed while providing a 7.8x larger per-layer receptive field. Combined with inter-block cross-window communication and hierarchical window reparameterization, WBMM achieves comparable or higher accuracy on ImageNet-1K, COCO, and ADE20K with 1.31-1.88x training speedup, and demonstrates consistent advantages across GPU, CPU, and edge devices without requiring specialized acceleration kernels. Our code is available at https://github.com/wansong-s/WBMM.

cs.CV↗
arXiv2026

Stability of the smooth Casselman-Jacquet functor

We establish and prove several results for the smooth Casselman-Jacquet submodule and quotient functors for real reductive groups. In addition to exactness, surjectivity, transitivity and globalization results, we establish a stability property for the intersection of Jacquet subspaces. Our approach is based on a family of seminorms on the Casselman-Jacquet quotient module. As an application, we establish a full version of the real Bernstein-Zelevinsky filtrations for smooth Fréchet representations of moderate growth.

math.RT↗
arXiv2026

Test-Time Curriculum for Open-Set AIGC Detection

AI-generated image detectors deployed in open-world environments inevitably face distribution shifts as new and stronger generative models continue to emerge. Although existing methods improve cross-generator generalization through better representations or training data construction, they typically follow a static train-once-and-deploy paradigm and cannot adapt after deployment. In this work, we study open-set AIGC image detection from a test-time adaptation perspective. We propose Test-Time Curriculum (TTC), a simple and model-agnostic framework that adapts a detector on unlabeled test data through curriculum-based self-training. TTC starts from highly reliable pseudo-labeled samples and progressively incorporates harder yet informative cases, while enforcing class-balanced selection to reduce biased updates under generator shift. To further improve pseudo-label quality, we introduce Cross-Scale Pseudo-Label Refinement, which aggregates complementary evidence across multiple resolutions for more reliable adaptation, and applies noisy-or fusion at inference to strengthen final predictions. In addition, we construct AIGCGuard, a new benchmark containing 3,100 representative real images and 124,000 generated images from 40 of the most advanced open-source and proprietary text-to-image models. Extensive experiments on five benchmarks show that TTC substantially improves overall detection performance under diverse unseen-generator shifts, establishing a practical and effective test-time adaptation framework for open-set generated image detection.

cs.CV↗
arXiv2026

CXRAgent: Director-Orchestrated Multi-Stage Reasoning for Chest X-Ray Interpretation

Chest X-ray (CXR) plays a pivotal role in clinical diagnosis, and a variety of task-specific and foundation models have been developed for automatic CXR interpretation. However, these models often struggle to adapt to new diagnostic tasks and complex reasoning scenarios. Recently, LLM-based agent models have emerged as a promising paradigm for CXR analysis, enhancing model's capability through tool coordination, multi-step reasoning, and team collaboration, etc. However, existing agents often rely on a single diagnostic pipeline and lack mechanisms for assessing tools' reliability, limiting their adaptability and credibility. To this end, we propose CXRAgent, a director-orchestrated, multi-stage agent for CXR interpretation, where a central director coordinates the following stages: (1) Tool Invocation: The agent strategically orchestrates a set of CXR-analysis tools, with outputs normalized and verified by the Evidence-driven Validator (EDV), which grounds diagnostic outputs with visual evidence to support reliable downstream diagnosis; (2) Diagnostic Planning: Guided by task requirements and intermediate findings, the agent formulates a targeted diagnostic plan. It then assembles an expert team accordingly, defining member roles and coordinating their interactions to enable adaptive and collaborative reasoning; (3) Collaborative Decision-making: The agent integrates insights from the expert team with accumulated contextual memories, synthesizing them into an evidence-backed diagnostic conclusion. Experiments on various CXR interpretation tasks show that CXRAgent delivers strong performance, providing visual evidence and generalizes well to clinical tasks of different complexity. Code and data are valuable at this \href{https://github.com/laojiahuo2003/CXRAgent/}{link}.

cs.AI↗
arXiv2026

MG$^2$-RAG: Multi-Granularity Graph for Multimodal Retrieval-Augmented Generation

Retrieval-Augmented Generation (RAG) mitigates hallucinations in Multimodal Large Language Models (MLLMs), yet existing systems struggle with complex cross-modal reasoning. Flat vector retrieval often ignores structural dependencies, while current graph-based methods rely on costly ``translation-to-text'' pipelines that discard fine-grained visual information. To address these limitations, we propose \textbf{MG$^2$-RAG}, a lightweight \textbf{M}ulti-\textbf{G}ranularity \textbf{G}raph \textbf{RAG} framework that jointly improves graph construction, modality fusion, and cross-modal retrieval. MG$^2$-RAG constructs a hierarchical multimodal knowledge graph by combining lightweight textual parsing with entity-driven visual grounding, enabling textual entities and visual regions to be fused into unified multimodal nodes that preserve atomic evidence. Building on this representation, we introduce a multi-granularity graph retrieval mechanism that aggregates dense similarities and propagates relevance across the graph to support structured multi-hop reasoning. Extensive experiments across four representative multimodal tasks (i.e., retrieval, knowledge-based VQA, reasoning, and classification) demonstrate that MG$^2$-RAG consistently achieves state-of-the-art performance while reducing graph construction overhead with an average 43.3$\times$ speedup and 23.9$\times$ cost reduction compared with advanced graph-based frameworks.

cs.IR↗
arXiv2026

BDFlow-3DRM: Height-Coherent 3D Radio Map Construction via Bi-Dynamical Flow Matching

Three-dimensional (3D) radio map (RM) is a key enabler for environment-aware communications in low-altitude wireless scenarios by providing site-specific channel priors indexed by spatial locations. However, existing 3D RM construction methods lack effective modeling of the height dimension, which limits their generalization to unseen height configurations and degrades construction coherence across height layers. In this paper, we propose BDFlow-3DRM, a bi-dynamical flow matching framework for 3D RM construction. Specifically, the 3D RM construction problem is formulated as a deterministic probability flow in a semantic latent space. BDFlow-3DRM learns continuous height-aware representations from flexible transceiver height inputs, enhancing geometric awareness. Meanwhile, its bi-dynamical design explicitly models bidirectional dependencies across neighboring height layers, so that RMs at different height layers can be constructed jointly. Extensive experiments on multiple datasets, covering diverse simplified and realistic urban scenarios, validate the effectiveness of BDFlow-3DRM. Compared with diffusion-based baselines, it reduces the normalized mean square error (NMSE) by 28.6% and attains a 180-fold reduction in inference complexity. More importantly, with only 20 training receiver-height layers, BDFlow-3DRM maintains accurate prediction over a wide continuous receiver-height range from 1 to 120 m under variable transmitter heights, highlighting its practical potential for large-scale 3D RM construction.

eess.SP↗
arXiv2026

GeNeRT: A Physics-Informed Approach to Intelligent Wireless Channel Modeling via Generalizable Neural Ray Tracing

Neural ray tracing (RT) has emerged as a promising paradigm for channel modeling by integrating physical propagation principles with neural networks. However, existing neural RT methods remain limited by strong spatial dependence and weak adherence to electromagnetic laws. We propose GeNeRT, a generalizable neural RT framework that improves generalization and accuracy through relative geometric features, scatterer semantics, and a Fresnel-inspired polarization-driven architecture. GeNeRT is trained through a three-stage strategy: polarization-specific module-wise pre-training captures general ray-surface interaction behavior; system-wise end-to-end training uses only receiver-side channel impulse responses to learn site-specific propagation characteristics; and measurement-based fine-tuning employs sparse measured multipath components (MPCs) to adapt polarization-related modules to real-world environments. Extensive outdoor simulations demonstrate robust intra-scenario transferability and inter-scenario zero-shot generalization. In an unseen scenario, GeNeRT achieves an overall error of $-35.36$ dB and an average-delay error of 4.91 ns, compared with $-10.85$ dB and 32.38 ns for the best baseline. With only 75 measured reflected MPCs, fine-tuning further reduces the overall error from $-14.48$ to $-22.90$ dB and the average-delay error from 6.28 to 3.58 ns. Ablation studies confirm the effectiveness of the proposed architecture and training strategy.

cs.LG↗
arXiv2026

Canonical Bernstein-Zelevinsky Filtration and Casselman's Comparison Conjecture

We establish a canonical Bernstein--Zelevinsky filtration for Casselman--Wallach representations that is analogous to the $p$-adic case. In addition, we outline an approach to Casselman's comparison conjecture and prove it for general linear groups, as well as for quasi-split even orthogonal groups in some special cases. We also give some applications of the Bernstein--Zelevinsky filtration, such as to the study of highest derivatives and the indecomposability of mirabolic restrictions.

math.RT↗
arXiv2026

Minimum free energy randomized design to improve covariate balance

``Block what you can and randomize what you cannot'' is the core principle for treatment effect estimation in randomized controlled trials. Although a wealth of allocation strategies has been developed, an explicit trade-off between the covariate balance achieved by blocking and the robustness guaranteed by randomization is seldom quantified. Motivated by the second law of thermodynamics, this work posits a new criterion that lowers the covariate imbalance while maximizing the entropy that quantifies contrast and allocation diversity. The resulting optimal strategy, termed the minimum free energy randomized design, is then derived, thereby formally achieving such a trade-off. To facilitate practical implementation, we further develop a computationally efficient dynamic allocation algorithm with theoretical guarantees. Using a finite-sample variance decomposition, the proposed randomization strategy is shown to control covariate imbalance while preventing unobserved heterogeneity from dominating the mean squared error, thus retaining minimax efficiency under the prescribed design constraints. Extensive numerical simulations demonstrate that our method achieves superior statistical efficiency and greater robustness than existing approaches.

stat.ME↗
arXiv2026

Fourier positivity for spherical functions I: split tori and spherical principal series

We prove Fourier positivity for spherical functions on a semisimple linear algebraic group $G$ over a local field restricted to its split tori $A$ for unitary principal series parameters of $G$. For ${\rm SL}_n(F)$, where $F$ is a local field, we obtain an explicit recursive formula for the Fourier transform on the diagonal split torus in terms of local Rankin--Selberg factors for ${\rm GL}_n\times {\rm GL}_{n-1}$, together with uniform exponential lower bounds in the spectral parameters. The main input is a Plancherel expansion for the restriction of a ${\rm GL}_n(F)$-spherical function to ${\rm GL}_{n-1}(F)$. Its coefficients are spherical periods computed by Rankin--Selberg theory. Positivity of the Fourier transform for general semisimple groups with unitary principal series parameters is obtained by reduction to full-rank subgroups of type A. The results are motivated by variance non-vanishing problems for mixing abelian actions on homogeneous spaces.

math.RT↗