Search arXiv⌕ Search

arXiv subjects

Jiaqi Yu

Publications and source records attributed to Jiaqi Yu.

At least 19 recordsLinked to original sources
arXiv2026

OcTreeMamba: Topology-Aware Mamba for 3D Medical Segmentation via Octree Sparsification

Accurate segmentation of 3D anatomical structures is crucial to medical image analysis. Recently, Mamba-based approaches have achieved promising segmentation performance in capturing long-range dependencies with linear-time computational complexity. However, heuristic selective scanning strategies inherently disrupt the intrinsic spatial hierarchy and increase computational redundancy, ignoring the inherent sparsity of 3D anatomical structures. To this end, we propose OcTreeMamba, a hierarchical framework that incorporates Octree structures into SSM, achieving 3D topology-aware long-range context feature propagation and integration. Specifically, a Structure Learner performs anatomy-adaptive structured sparsification through hierarchical gating and Gumbel-Softmax sampling, enabling differentiable selection of anatomical regions. The selected regions are subsequently organized by a Dynamic Octree Builder into a multi-level topology that preserves their hierarchical connectivity. Based on the constructed Octree, a bidirectional Tree-SSM propagates and integrates contextual features along connectivity-aware leaf--parent--root and root--parent--leaf trajectories. Extensive experiments on four benchmark datasets show that our proposed OcTreeMamba achieves competitive performance \wrt topological continuity and computational efficiency, particularly for recovering the complex structure on vascular datasets. The code is available at https://github.com/MikeMak123/OcTreeMamba.

cs.CV↗
arXiv2026

Observability and stabilizability for Grushin-Type parabolic equations in the whole space

We study observability and stabilization for the whole-space Grushin heat equation $$ \partial_t u-Δ_γu=0,\qquad Δ_γ:=Δ_x+|x|^{2γ}Δ_y,\qquad (t,x,y)\in(0,T)\times\mathbb{R}^n\times\mathbb{R}^m,\quad γ>0. $$ For product sets $E=E_1\times E_2$, we identify three observability regimes in which thickness, distance from the degeneracy set $\{x=0\}$, and local density near that set play different roles. (i) In the subcritical regime $0<γ<1$, the observability inequality holds for every $T>0$ if and only if both $E_1$ and $E_2$ are thick; hence $T_{\min}(E)=0$. (ii) At the critical exponent $γ=1$, if $E_2=\mathbb{R}^m$ and $E_1\subset\mathbb{R}^n$ is thick, we determine the exact minimal observable time: $$ T_{\min}(E)=\frac{a_{\mathrm{ess}}(E_1)^2}{2n},\qquad a_{\mathrm{ess}}(E_1):=\operatorname*{ess\,inf}_{x\in E_1}|x|. $$ Thus, the threshold depends on the essential distance of $E_1$ from the degeneracy set. (iii) In the supercritical regime $γ>1$, we consider $n=1$, $E_2=\mathbb{R}^m$ and thick sets $E_1\subset\mathbb{R}$. If $a_{\mathrm{ess}}(E_1)>0$, observability fails for every $T>0$, so that $T_{\min}(E)=+\infty$. If $a_{\mathrm{ess}}(E_1)=0$, observability depends on the density of $E_1$ near $0$, measured by $$ D_{E_1}(r):=\log\!\left(\frac{2r}{|E_1\cap(-r,r)|}\right),\qquad L_{E_1}:=\limsup_{r\downarrow0}r^2D_{E_1}(r). $$ If $L_{E_1}=0$, then $T_{\min}(E)=0$; if $0 0$, exponential and complete stabilizability hold if and only if both $E_1$ and $E_2$ are thick. The stabilization criterion is therefore independent of $γ$.

math.AP↗
arXiv2026

LDDR: Linear-DPP Dynamic-Resolution Frame Sampling for Video MLLMs

Multimodal large language models, when used for long video understanding, require selecting an informative frame set from massive and redundant candidates. Limited by the model's visual-token budget and reasoning ability, frame selection is therefore essential to reduce input size while preserving the most relevant information for accurate video reasoning. Existing frame-sampling methods like uniform sampling or relevance/diversity-sampling often miss important events or retain redundant frames. On the other hand, agentic frame selection methods incur very high computational costs. Further, these methods allocate equal tokens to all selected frames, regardless of their importance. We therefore propose LDDR (Linear DPP Dynamic Resolution), which is a training-free and budget-aware video frame sampling framework. It is computationally efficient and allows non-uniform token budget allocation within the candidate frame set. Specifically, LDDR performs query-aware Determinantal Point Process (DPP) frame selection in a task-conditioned feature space, achieving a 3x runtime speedup over standard DPP baselines. It further introduces a Group DPP importance score to guide frame retention and dynamic resolution allocation: assigning more tokens to informative and non-redundant frames while downscaling or pruning less useful ones. We test LDDR on four popular video benchmarks spanning short, medium, and long-range videos. LDDR achieves a 2.5-point gain in low-budget settings and a 1.6-point gain in high-budget scenarios. These improvements are consistently observed across multiple MLLM families and sizes. Further manual qualitative analyses are included to prove the method's effectiveness.

cs.CV↗
arXiv2026

Spectral Rendering Without a Spectral Renderer: Learned Spectral Codes for RGB Pipelines

Spectral rendering reproduces the wavelength-dependent appearance that RGB rendering cannot: metamerism, colour shifts under spectrally rich illumination, coloured glass, metallic reflection and subsurface scattering. It requires a spectrally aware renderer, but the rendering ecosystem is predominantly RGB: game engines, cloud render farms, GPU rasterisers and most offline renderers consume RGB alone, with no access to the internal wavelength propagation that spectral rendering relies on. We introduce learned spectral codes, a compact linear representation that enables general spectral rendering through unmodified RGB pipelines. A trained encoder compresses any visible-range spectral quantity into a small latent vector; the codes are packed as RGB triplets and processed by an off-the-shelf RGB renderer. The codec preserves the arithmetic of light transport, exactly for scaling and addition and approximately for element-wise products, so two standard RGB passes suffice to produce results closely matching a full spectral rendering. The same codec handles reflectance, illumination, coloured glass, conductors, and participating media: one pipeline across material classes, with no per-material modification to the renderer. Where spectral asset data is unavailable, lightweight neural upsamplers bridge legacy RGB inputs to the latent space. Across diverse materials and illumination, the method yields substantial quality improvements on the cases where RGB rendering fails most visibly, at rendering cost comparable to efficient spectral-sampling techniques and with no changes to the renderer.

cs.GR↗
arXiv2026

A c-transform optimal transport method for high-contrast freeform reflector design

Design of high-contrast freeform reflectors is challenging, as the presence of zero-intensity regions leads to degeneracy in the associated Monge-Ampere-type equation. A common remedy is to add a positive artificial background to the target intensity, which improves the regularity of the equation. However, this regularization inevitably reduces the achievable illumination contrast and introduces nonzero intensity into regions that are intended to remain dark. We propose a fast c-transform method based on optimal transport duality for high-contrast freeform reflector design in the far field without artificial background regularization. The method integrates repeated discrete c-transforms into a measure-based dual optimization scheme to maintain the c-concavity of the reflector potential throughout the iteration. This preserves the admissibility of the iterates and enables stable convergence for targets with zero-intensity regions. To efficiently compute the discrete c-transforms, we develop a localized search algorithm that exploits their contact structure and prove its exactness and linear complexity. Numerical experiments demonstrate that the proposed method accurately realizes high-contrast illumination targets with zero background.

math-ph↗
arXiv2026

Quantitative observability of the Schrödinger equation in the torus from open sets

We study the observability of the Schrödinger equation on the $d$-dimensional torus $\T^d=\R^d/(2π\Z)^d$, $d \geq 1$, from an open subset $ω\subset \T^d$. Our main result establishes a quantitative observability estimate for the free Schrödinger equation in the regime of small times $T$ and small observation sets of the form $ω= \prod_{j=1}^{d}(a_j,b_j)$. A central ingredient in the proof is a cluster decomposition method combined with an induction scheme introduced by Bourgain and further developed by Burq and Zhu.

math.AP↗
arXiv2026

Offline-Online Curriculum RL for Multimodal Reasoning

Multimodal large language models exhibit capabilities on reasoning tasks, yet often produce flawed intermediate steps while yielding correct final answers. This behavior undermines interpretability and reliability, suggesting reliance on spurious shortcuts rather than faithful reasoning. Although efforts have explored step-level supervision, distinguishing decisive steps from redundant ones remains challenging. We propose $O^2$-CritiCuRL, a novel curriculum reinforcement learning framework that introduces critical-step awareness through an iterative offline-online paradigm. In the offline stage, $O^2$-CritiCuRL conducts multi-rollout analysis over step-annotated trajectories to estimate step-level importance, allowing the framework to distill critical reasoning steps and filter out redundant ones. In the online stage, we employ a progressive step-level reinforcement learning strategy, where truncated chains guide the model to infer missing steps and refine its reasoning, thereby sharpening its focus on critical steps and overcoming the limitations of static supervision. Extensive experiments on multimodal reasoning benchmarks show that our method achieves state-of-the-art performance while delivering superior training and inference efficiency. Code is available at https://github.com/kk0013/CritiCuRL.

cs.AI↗
arXiv2026

Fractal Turán-Nazarov Inequality and Observability for Schrödinger Equations

This paper establishes limitations on observability inequality and unique continuation for Schrödinger equations on fractal sets. We prove that, in contrast to the heat equation, such properties can fail in fractal settings. To achieve this, we first extend the classical Turán--Nazarov inequality, which provides lower bounds of trigonometric polynomials of the form $\sum_{k=1}^nc_ke^{2πim_kt}$ on sets of positive measure, to the fractal setting. Unlike in the classical case, the constant in the inequality loses uniformity in the degree $n$, and we obtain sharp bounds depending on both $n$ and the frequency difference $m_n-m_1$. These refinements then enable us to construct explicit counterexamples, showing that observability and unique continuation may fail for Schrödinger equations when the observation set is fractal.

math.AP↗
arXiv2026

Uncertainty as Remedy: Mitigating Satisfaction Label Bias in Short Video Multi-Objective Ensemble Ranking

The core objective of short video recommendation is to model users' unobservable true satisfaction with recommended videos. As the dominant industrial framework, end-to-end multi-objective ensemble ranking models are typically trained with multi-dimensional dense user behavioral signals, such as clicks and watch time. However, these behavioral signals are partial, fragmented, and often mutually conflicting user satisfaction proxies, introducing uncertainty and label bias into satisfaction modeling. Conventional deterministic models overlook this uncertainty, which exacerbates satisfaction label bias and results in suboptimal model convergence. Meanwhile, existing uncertainty-aware methods mostly employ uncertainty for post-hoc ranking adjustments rather than leveraging it as a remedy to mitigate the inherent bias within the core optimization pipeline. This paper proposes UAME, an Uncertainty-Aware end-to-end Multi-objective Ensemble ranking framework for short video recommendation. UAME represents the model's prediction as a Gaussian scoring variable, where the mean denotes the predicted satisfaction score and the variance quantifies predictive uncertainty associated with this score. We further design a probabilistic pairwise ranking loss, and construct an uncertainty-aware sample-level weighting scheme to mitigate the bias. We further provide theoretical analysis suggesting that the weighting scheme helps mitigate satisfaction label bias. Extensive offline and online experiments on a large-scale industrial short video platform demonstrate that UAME consistently improves two state-of-the-art paradigms, EMER and EASQ, and better aligns with questionnaire-based user satisfaction. UAME has been deployed in our production short-video recommendation system and continues to deliver stable, statistically significant gains.

cs.IR↗
arXiv2026

SentGuard: Sentence-Level Streaming Guardrails for Large Language Models

Large language models increasingly stream long, reasoning-intensive responses in real time, making when to moderate as critical as whether to moderate. Existing guardrails fall into two unsatisfactory extremes: response-level methods delay intervention until the full output is generated, whereas token-level methods act on incomplete semantics, often producing unstable decisions and excessive guard invocations. To address this challenge, we propose SentGuard, a sentence-level streaming guardrail that operates in parallel with generation. A lightweight waiting buffer groups streamed tokens into sentence chunks and releases only verified chunks to the user, introducing a small offset that enables SentGuard to assess the current prefix while the target LLM decodes subsequent content. To support this, we construct StreamSafe, a benchmark with structured per-sentence annotations across 8 harm categories, capturing the evolution of safety risks across both reasoning and response segments. We further train SentGuard with a coarse-to-fine objective to detect unsafe intent as soon as it emerges at sentence boundaries. Experiments on 5 safety benchmarks show that SentGuard outperforms existing baselines, detecting 90.5% of unsafe cases within two sentences while maintaining a low streaming false-positive rate of 7.41%.

cs.CL↗
arXiv2026

Saturation for product systems of polynomials

Recently, Qiu, Xu, Ye and Yu proved that for product system of finitely many minimal systems, the maximal $\infty$-step pro-nilfactor of the system is the topological characteristic factor. In this paper, we extend the result to polynomials.

math.DS↗
arXiv2026

S2Aligner: Pair-Efficient and Transferable Pre-Training for Sparse Text-Attributed Graphs

Pre-training on text-attributed graphs (TAGs) is central to building transferable graph foundation models, where LLM-as-Aligner methods align graph and text representations through the semantic knowledge of large language models. However, these methods usually assume that node texts provide sufficient and reliable supervision, an assumption often violated in real-world sparse TAGs. When textual anchors are missing, noisy, or uneven across domains, graph structures must be aligned with weak semantic evidence, leading to unreliable structure-semantics correspondence and sparsity-induced transfer bias. This paper presents S2Aligner, a sparsity-aware and structure-enhanced LLM-as-Aligner framework for graph-text pre-training on sparse TAGs. The key idea is to decouple semantic alignment from structural modeling, allowing topology-aware signals to enhance alignment without contaminating the shared semantic space. Specifically, S2Aligner decomposes graph-text representations into semantic and structural components, uses structure-oriented reconstruction with consistency control to inject reliable topology cues into text representations, and suppresses inconsistent structural signals under textual sparsity. Moreover, S2Aligner introduces sparsity-aware cross-domain risk balancing, which calibrates domain risks through a global-domain density ratio and downweights unreliable sparse samples via graph reliability estimation. Theoretical analysis shows that this objective reduces cross-domain generalization gaps by controlling domain risk discrepancy. Extensive experiments across diverse graph domains, sparsity levels, and downstream tasks demonstrate that S2Aligner consistently outperforms existing baselines.

cs.LG↗
arXiv2026

TAME: Test-Time Adversarial Prompt Tuning via Mixture-of-Experts for Vision-Language Models

Large-scale pre-trained Vision-Language models (VLMs), such as CLIP, exhibit strong zero-shot generalization, yet remain highly vulnerable to imperceptible adversarial perturbations, raising serious safety concerns for open-world deployment. To enhance robustness without requiring downstream task-specific retraining, we propose TAME, a novel test-time defense. Building upon our prior Test-Time Adversarial Prompt Tuning (TAPT), TAME introduces an architectural reformulation by replacing TAPT's single adaptive prompt with an input-conditioned Mixture-of-Experts (MoE) framework, enabling more expressive and adaptive defense. Specifically, TAME maintains a bank of learnable expert prompts and employs an input-dependent routing mechanism to aggregate a customized prompt mixture for each unlabeled test sample at inference time. This test-time defense mechanism is driven by three unsupervised objectives: (1) multi-view prediction entropy minimization, (2) layer-wise alignment of visual token statistics to precomputed clean and adversarial reference distributions, and (3) MoE regularization for balanced expert utilization and prompt diversity. We evaluated TAME on 11 benchmark datasets, including ImageNet and 10 additional zero-shot datasets. The results show that TAME improves the zero-shot adversarial robustness of the original CLIP by at least 49.1% under AutoAttack while largely preserving generalization on clean samples. TAME also consistently outperforms existing adversarial prompt tuning methods across multiple prompt designs, yielding an average robustness gain of at least 30.2%.

cs.CV↗
arXiv2026

EPAR: Electromagnetic Pathways to Architectural Reliability in Quantum Processors

As superconducting processors scale, understanding how physical layout shapes qubit interactions is essential for architectural reliability. Existing methods offer limited insight into how electromagnetic design choices translate into execution-level behavior. We present EPAR, an electromagnetic-to-architecture framework that predicts robustness early directly from physical design by reconstructing how design distortion modifies the effective Hamiltonian, reroutes mediated connectivity, and influences control-pulse response. Across all tested layouts, EPAR's structural scores show 100% agreement with two-qubit error trends yet reveal over 10X robustness differences among edges with identical calibrated error rates, going beyond conventional metrics to provide improved and actionable compiler guidance.

cs.ET↗
arXiv2026

Towards End-to-End Alignment of User Satisfaction via Questionnaire in Video Recommendation

Short-video recommender systems typically optimize ranking models using dense user behavioral signals, such as clicks and watch time. However, these signals are only indirect proxies of user satisfaction and often suffer from noise and bias. Recently, explicit satisfaction feedback collected through questionnaires has emerged as a high-quality direct alignment supervision, but is extremely sparse and easily overwhelmed by abundant behavioral data, making it difficult to incorporate into online recommendation models. To address these challenges, we propose a novel framework which is towards End-to-End Alignment of user Satisfaction via Questionaire, named EASQ, to enable real-time alignment of ranking models with true user satisfaction. Specifically, we first construct an independent parameter pathway for sparse questionnaire signals by combining a multi-task architecture and a lightweight LoRA module. The multi-task design separates sparse satisfaction supervision from dense behavioral signals, preventing the former from being overwhelmed. The LoRA module pre-inject these preferences in a parameter-isolated manner, ensuring stability in the backbone while optimizing user satisfaction. Furthermore, we employ a DPO-based optimization objective tailored for online learning, which aligns the main model outputs with sparse satisfaction signals in real time. This design enables end-to-end online learning, allowing the model to continuously adapt to new questionnaire feedback while maintaining the stability and effectiveness of the backbone. Extensive offline experiments and large-scale online A/B tests demonstrate that EASQ consistently improves user satisfaction metrics across multiple scenarios. EASQ has been successfully deployed in a production short-video recommendation system, delivering significant and stable business gains.

cs.IR↗
arXiv2025

An End-to-End Multi-objective Ensemble Ranking Framework for Video Recommendation

We propose a novel End-to-end Multi-objective Ensemble Ranking framework (EMER) for the multi-objective ensemble ranking module, which is the most critical component of the short video recommendation system. EMER enhances personalization by replacing manually-designed heuristic formulas with an end-to-end modeling paradigm. EMER introduces a meticulously designed loss function to address the fundamental challenge of defining effective supervision for ensemble ranking, where no single ground-truth signal can fully capture user satisfaction. Moreover, EMER introduces novel sample organization method and transformer-based network architecture to capture the comparative relationships among candidates, which are critical for effective ranking. Additionally, we have proposed an offline-online consistent evaluation system to enhance the efficiency of offline model optimization, which is an established yet persistent challenge within the multi-objective ranking domain in industry. Abundant empirical tests are conducted on a real industrial dataset, and the results well demonstrate the effectiveness of our proposed framework. In addition, our framework has been deployed in the primary scenarios of Kuaishou, a short video recommendation platform with hundreds of millions of daily active users, achieving a 1.39% increase in overall App Stay Time and a 0.196% increase in 7-day user Lifetime(LT7), which are substantial improvements.

cs.IR↗
arXiv2024

DREAM: Domain-agnostic Reverse Engineering Attributes of Black-box Model

Deep learning models are usually black boxes when deployed on machine learning platforms. Prior works have shown that the attributes (e.g., the number of convolutional layers) of a target black-box model can be exposed through a sequence of queries. There is a crucial limitation: these works assume the training dataset of the target model is known beforehand and leverage this dataset for model attribute attack. However, it is difficult to access the training dataset of the target black-box model in reality. Therefore, whether the attributes of a target black-box model could be still revealed in this case is doubtful. In this paper, we investigate a new problem of black-box reverse engineering, without requiring the availability of the target model's training dataset. We put forward a general and principled framework DREAM, by casting this problem as out-of-distribution (OOD) generalization. In this way, we can learn a domain-agnostic meta-model to infer the attributes of the target black-box model with unknown training data. This makes our method one of the kinds that can gracefully apply to an arbitrary domain for model attribute reverse engineering with strong generalization ability. Extensive experimental results demonstrate the superiority of our proposed method over the baselines.

cs.LG↗