Searcharxiv⌕ Search

arXiv subjects

Yufeng Zhang

Publications and source records attributed to Yufeng Zhang.

At least 19 recordsLinked to original sources
arXiv2026

Statistical behavior of systems with nested invariant cones

Nested invariant cones (NICs) encode hierarchical order and oscillation structures in many finite and infinite dimensional dynamical systems. We investigate how this hierarchy constrains the Birkhoff center and the supports of invariant probability measures. For eventually compact, dissipative semiflows that are uniformly eventually strongly monotone with respect to NICs, we prove that every connected component $B$ of the Birkhoff center lies in a single cone layer: there exists $j>0$ such that $B$ is unordered with respect to all lower-level cones and strongly ordered with respect to all cones at level $j$ or higher. The same conclusion holds for every connected component of the support of an invariant probability measure. Under an additional transversality condition involving a codimension-$d$ linear subspace, each such component admits a homeomorphic embedding into $\mathbb R^d$. If $d=1$ or $2$ and the restriction of the semiflow to its global attractor extends to a flow, then the system has zero topological entropy, independently of the dimension of the original phase space. We apply the theory to bidirectional cyclic feedback systems and scalar parabolic equations on the circle. In the parabolic case, the natural infinite family of zero-number NICs fails to be uniformly eventually strongly monotone, even for the heat equation; we overcome this obstruction by constructing a finite family of perturbed NICs. In both applications, every connected component of the Birkhoff center admits a planar embedding, and the topological entropy is zero.

math.DS↗
arXiv2026

From Monolithic Blending to Agentic Orchestration: Dynamic Response for Conversational Assistants at Scale

Conversational assistants can blend retrieval, action selection, escalation, and wording in a single model path, or separate those roles. We report a production migration of a customer-support assistant at a large accommodation marketplace (millions of conversations per month, 11 languages, 10-second P90). Dynamic Response (DR) replaces a single Qwen3-235B-A22B blended responder with a bounded ReAct orchestrator over typed tools plus a smaller generator that writes from a backend-validated context contract. Because the migration also changed prompts, alignment, and serving, we attribute each effect to its cause and claim as architecture effects only those measured on identical replayed turns: typed entity selection moves the reservation selector to a precision-first operating point (precision 8.3% to 89.1%, recall 75.2% to 67.3%), and typed action IDs with a membership check remove observed structured-action hallucination (2.14% to 0.0%). A low-ramp A/B test reproduces the replay escalation reductions: hard-escalation responses fall from 5.60% to 3.08% and soft-escalation responses from 9.56% to 2.49%, while production handoff volume holds roughly steady; self-solve is directional (+5.1 points, 95% CI [-2, +12]). Serving optimizations cut orchestrator P90 latency from 3.87s to 2.24s on a GPU footprint reduced by roughly one-third, and self-hosting reduces estimated annual model-serving cost by more than an order of magnitude.

cs.AI↗
arXiv2026

Artificial Intelligence in a Photonic Temporal Processor

Optical neural networks (ONNs) promise high-throughput and energy-efficient artificial intelligence, yet essentially all implementations so far encode information across space either in free-space arrays or in integrated waveguide meshes, tying the number of neurons to the number of physical components and fixes the routing topology at fabrication. Here we show that moving the computation into time decouples computational dimension from hardware dimension. Exploiting space-time duality, we implement optical diffraction and interference entirely in time domain, using thin-film lithium niobate modulators as time lenses and temporal masks, with chromatic dispersion providing the coupling between successive temporal neurons. We experimentally verify high-order, complex-valued matrix-matrix multiplications using just a single optical input/output port, scaling the computational dimensions far beyond the channel count. By incorporating optical feedback, we extend this platform into versatile neural networks, where the network layers, neuron numbers, and synaptic connections are fully programmable and in-situ trainable. Our temporal diffractive neural networks are successfully validated on various classification benchmarks, alongside image and video generation tasks. Notably, using this platform we demonstrate an all-analogue generative pipeline in which the latent variable is drawn directly from amplified spontaneous emission, so that no digital sampling or electronic modulation appears anywhere in the generative path. Furthermore, high-resolution images and videos are generated at high frame rates, outperforming state-of-the-art modulator-refresh-limited optical generative systems. These results establish a unified photonic temporal computing framework, providing a scalable and deployable pathway toward next-generation machine intelligence.

physics.optics↗
arXiv2026

Birkhoff center and recurrent behavior of differentially positive systems on a homogeneous space

We study the structure of the Birkhoff center and the recurrent behavior of differentially positive systems whose linearizations along trajectories preserve a homogeneous cone field on a homogeneous space. Such cone fields arise naturally from general relativity and Lie theory. We establish an order-structural dichotomy for the Birkhoff center: every connected component of the Birkhoff center, as well as the support of any invariant measure, is either strongly ordered or unordered. This yields a comprehensive characterization of recurrent dynamics for differentially positive systems on homogeneous spaces, interpreted through the lens of the underlying order relation.

math.DS↗
arXiv2026

CORA-Diff: Confidence-Oriented Residual Acceptance for Efficient Diffusion Language Model Inference

Diffusion language models (DLMs) update many tokens in parallel, yet practical decoders often use a fixed denoising horizon. Many predictions stabilize early, but blockwise decoding continues until all positions are resolved, causing repeated dense forward passes. Existing accelerators often rely on learned filters, modified scores, dependency models, or cache-specific mechanisms. We ask whether native trajectory signals can identify residual positions likely to match the deterministic dense endpoint. We propose CORA-Diff, a training-free method that preserves the original transfer rule and applies confidence-and-persistence gating only to positions that rule leaves unresolved. Accepted tokens remain visible as context, and the block terminates once all positions are resolved. This requires no backbone change, learned acceptance model, or logit modification. Our theory explains why high-confidence, persistent predictions are more likely to match the fixed-horizon dense endpoint, and paired post-intervention trajectories provide direct empirical support. We select one operating point on a separate GSM8K calibration subset and freeze it for all evaluations. Under a matched Learn2PD-style LLaDA protocol, CORA-Diff has the lowest measured runtime in all eight task-length settings. Task scores match or exceed dense decoding in five settings, and the largest observed drop is 1.22 points. Its incremental speedups over EOS-aware dense decoding are 2.70x and 3.32x on GSM8K and HumanEval. It also reaches 13.14x under the fixed-horizon 1024/1024 mechanism-isolation protocol and transfers to Dream without retuning at 3.18x-3.53x. These results show that native confidence and persistence enable reliable residual acceptance, reducing repeated denoising computation while preserving task quality.

cs.AI↗
arXiv2026

Arbitrary Reflectionless Optical Routing via Non-Hermitian Zero-Index Networks

Optical routers are fundamental to photonic systems, but their performance is often limited by unwanted reflections and constrained functionalities. Existing design strategies generally lack complete control over reflectionless pathways and typically require computationally intensive iterative optimization. A general analytical framework for the inverse design of arbitrary reflectionless routing has remained unavailable. Here, we present an analytical inverse-design approach based on non-Hermitian zero-index networks, which enables arbitrary reflectionless routing for nearly any desired scattering response. By establishing a direct algebraic mapping between target scattering responses and the network's physical parameters, we transform the design process from iterative optimization into deterministic calculation. This approach enables the precise engineering of arbitrary reflectionless optical routing. We demonstrate its broad utility by designing devices from unicast and multicast routers with full amplitude and phase control to coherent beam combiners and spatial mode demultiplexers in four-port and six-port networks. Our work provides a systematic and analytical route to designing advanced light-control devices.

physics.optics↗
arXiv2026

Field-Aware RankMixer with Dual-Stream Bilinear Fusion for the Tencent UNI-REC Challenge

This paper presents our solution to the KDD Cup 2026 Tencent UNIREC Challenge. The task requires joint modeling of multi-domain user behavior sequences and non-sequential multi-field features for target-ad pCVR prediction. We develop a Field-Aware RankMixer (FA-RankMixer) with dual-stream bilinear fusion. The model first applies target-aware DIN modules to extract user interests from multiple behavior domains. It also models recent and earlier interests separately for the longest behavior sequence. The model then forms semantic tokens based on feature fields and behavior domains and uses RankMixer blocks for cross-token interaction. A shallow MLP stream complements the deep RankMixer stream, and a group-wise bilinear module fuses their representations. Our final solution ranks ninth on the official leaderboard. Our code is available at https://github.com/PixelCookie-zyf/TAAC-2026-SeRankMixer.

cs.LG↗
arXiv2026

Booster Lab: A Data-Centric Pipeline for Learning Deployable Humanoid Locomotion Policies

Humanoid robot motion learning requires not only task-oriented control policies but also physically feasible and natural behaviors that can be transferred to real robots. However, robot-feasible motion data are often scarce: raw human demonstrations may be incompatible with the robot morphology, open-source clips vary in quality, and simulation-collected robot trajectories still require feasibility checking. To address these challenges, we propose a data-centric training and deployment pipeline that integrates motion data curation, real-to-sim model adaptation, AMP-based reinforcement learning, and sim-to-real deployment. We validate the framework on the Booster T1 robot and further provide preliminary cross-platform validation on Booster K1.

cs.RO↗
arXiv2026

Almost sure convergence of differentially positive systems on a globally orderable manifold

Differentially positive systems are nonlinear systems whose linearization along trajectories preserves a cone field on a smooth Riemannian manifold. The structures of cone field come from general relativity and Lie theory. We prove that on a globally orderable manifold, the set of convergent points has full Riemann-Lebesgue measure, thus establishing almost sure convergence. This result thereby resolves a measure-theoretic form of the conjecture posed by Forni and Sepulchre in 2016 for such manifolds.

math.DS↗
arXiv2026

IMPACT-HOI: Supervisory Control for Onset-Anchored Partial HOI Event Construction

We present IMPACT-HOI, a mixed-initiative framework for annotating egocentric procedural video by constructing structured event graphs for Human-Object Interactions (HOI), motivated by the need for high-quality structured supervision for learning robot manipulation from human demonstration. IMPACT-HOI frames this task as the incremental resolution of a partially specified, onset-anchored event state. A trust-calibrated controller selects among direct queries, human-confirmed suggestions, and conservative completions based on empirical annotator behavior and evidence quality. A risk-bounded execution protocol, utilizing atomic rollback, ensures that human-confirmed decisions are preserved against conflicting automated updates. A user study with 9 participants shows a 13.5% reduction in manual annotation actions, a 46.67% event match rate, and zero confirmed-field violations under the studied protocol. The code will be made publicly available at https://github.com/541741106/IMPACT_HOI.

cs.CV↗
arXiv2026

Optimal Stability of KL Divergence under Gaussian Perturbations

We study the problem of characterizing the stability of Kullback-Leibler (KL) divergence under Gaussian perturbations beyond Gaussian families. Existing relaxed triangle inequalities for KL divergence critically rely on the assumption that all involved distributions are Gaussian, which limits their applicability in modern applications such as out-of-distribution (OOD) detection with flow-based generative models. In this paper, we remove this restriction by establishing a sharp stability bound between an arbitrary distribution and Gaussian families under mild moment conditions. Specifically, let $P$ be a distribution with finite second moment, and let $\mathcal{N}_1$ and $\mathcal{N}_2$ be multivariate Gaussian distributions. We show that if $KL(P||\mathcal{N}_1)$ is large and $KL(\mathcal{N}_1||\mathcal{N}_2)$ is at most $ε$, then $KL(P||\mathcal{N}_2) \ge KL(P||\mathcal{N}_1) - O(\sqrtε)$. Moreover, we prove that this $\sqrtε$ rate is optimal in general, even within the Gaussian family. This result reveals an intrinsic stability property of KL divergence under Gaussian perturbations, extending classical Gaussian-only relaxed triangle inequalities to general distributions. The result is non-trivial due to the asymmetry of KL divergence and the absence of a triangle inequality in general probability spaces. As an application, we provide a rigorous foundation for KL-based OOD analysis in flow-based models, removing strong Gaussian assumptions used in prior work. More broadly, our result enables KL-based reasoning in non-Gaussian settings arising in deep learning and reinforcement learning.

cs.LG↗
arXiv2026

Relaxed Triangle Inequality for Kullback-Leibler Divergence Between Multivariate Gaussian Distributions

The Kullback-Leibler (KL) divergence is not a proper distance metric and does not satisfy the triangle inequality, posing theoretical challenges in certain practical applications. Existing work has demonstrated that KL divergence between multivariate Gaussian distributions follows a relaxed triangle inequality. Given any three multivariate Gaussian distributions $\mathcal{N}_1, \mathcal{N}_2$, and $\mathcal{N}_3$, if $KL(\mathcal{N}_1, \mathcal{N}_2)\leq ε_1$ and $KL(\mathcal{N}_2, \mathcal{N}_3)\leq ε_2$, then $KL(\mathcal{N}_1, \mathcal{N}_3)< 3ε_1+3ε_2+2\sqrt{ε_1ε_2}+o(ε_1)+o(ε_2)$. However, the supremum of $KL(\mathcal{N}_1, \mathcal{N}_3)$ is still unknown. In this paper, we investigate the relaxed triangle inequality for the KL divergence between multivariate Gaussian distributions and give the supremum of $KL(\mathcal{N}_1, \mathcal{N}_3)$ as well as the conditions when the supremum can be attained. When $ε_1$ and $ε_2$ are small, the supremum is $ε_1+ε_2+2\sqrt{ε_1ε_2}+o(ε_1)+o(ε_2)$. Finally, we demonstrate several applications of our results in out-of-distribution detection with flow-based generative models and safe reinforcement learning.

stat.ML↗
arXiv2026

Cooperative-Competitive Team Play of Real-World Craft Robots

Multi-agent deep Reinforcement Learning (RL) has made significant progress in developing intelligent game-playing agents in recent years. However, the efficient training of collective robots using multi-agent RL and the transfer of learned policies to real-world applications remain open research questions. In this work, we first develop a comprehensive robotic system, including simulation, distributed learning framework, and physical robot components. We then propose and evaluate reinforcement learning techniques designed for efficient training of cooperative and competitive policies on this platform. To address the challenges of multi-agent sim-to-real transfer, we introduce Out of Distribution State Initialization (OODSI) to mitigate the impact of the sim-to-real gap. In the experiments, OODSI improves the Sim2Real performance by 20%. We demonstrate the effectiveness of our approach through experiments with a multi-robot car competitive game and a cooperative task in real-world settings.

cs.RO↗
arXiv2026

On-chip Multimode Opto-electronic Neural Network

Opto-electronic computing combines the complementary strengths of photonics and electronics to deliver ultrahigh computational throughput with high energy efficiency. However, its practical deployment for real-world applications has been limited by architectures that rely on delicate wavelength management or phase-sensitive coherent detection. Here, we demonstrate the first multimode opto-electronic neural network (MOENN) on a silicon-on-insulator platform. By utilizing orthogonal waveguide eigenmodes as independent information carriers, our architecture achieves robust single-wavelength computation that is inherently immune to spectral crosstalk and phase noise. The fabricated MOENN chip monolithically integrates all functional components, including input encoders, programmable mode-division fan-in/-out units, and most importantly, the nonlinear multimode activation functions. We report the system's versatility through in-situ training via a genetic algorithm, successfully resolving the nonlinear decision boundaries of a two-class dataset and achieving 92.1% accuracy on the Iris classification benchmark. Furthermore, we reconfigure the MOENN into a one-dimensional convolutional neural network, attaining an accuracy of 90.7% on the electrocardiogram-based emotion recognition task. This work establishes a new opto-electronic computing paradigm of simple control and excellent robustness, providing a compelling path toward scalable, deployable photonic intelligence.

physics.optics↗
arXiv2026

S$^2$F: Principled Hybrid Testing With Fuzzing, Symbolic Execution, and Sampling

Hybrid testing that integrates fuzzing, symbolic execution, and sampling has demonstrated superior testing efficiency compared to individual techniques. However, the state-of-the-art (SOTA) hybrid testing tools do not fully exploit the capabilities of symbolic execution and sampling in two key aspects. First, the SOTA hybrid testing tools employ tailored symbolic execution engines that tend to over-prune branches, leading to considerable time wasted waiting for seeds from the fuzzer and missing opportunities to discover crashes. Second, existing methods do not apply sampling to the appropriate branches and therefore cannot utilize the full capability of sampling. To address these two limitations, we propose a novel hybrid testing architecture that combines the precision of conventional symbolic execution with the scalability of tailored symbolic execution engines. Based on this architecture, we propose several principles for combining fuzzing, symbolic execution, and sampling. We implement our method in a hybrid testing tool S$^2$F. To evaluate its effectiveness, we conduct extensive experiments on 15 real-world programs. Experimental results demonstrate that S$^2$F outperforms the SOTA tool, achieving an average improvement of 6.14% in edge coverage and 32.6% in discovered crashes. Notably, our tool uncovers three previously unknown crashes in real-world programs.

cs.SE↗
arXiv2025

MCOP: Multi-UAV Collaborative Occupancy Prediction

Unmanned Aerial Vehicle (UAV) swarm systems necessitate efficient collaborative perception mechanisms for diverse operational scenarios. Current Bird's Eye View (BEV)-based approaches exhibit two main limitations: bounding-box representations fail to capture complete semantic and geometric information of the scene, and their performance significantly degrades when encountering undefined or occluded objects. To address these limitations, we propose a novel multi-UAV collaborative occupancy prediction framework. Our framework effectively preserves 3D spatial structures and semantics through integrating a Spatial-Aware Feature Encoder and Cross-Agent Feature Integration. To enhance efficiency, we further introduce Altitude-Aware Feature Reduction to compactly represent scene information, along with a Dual-Mask Perceptual Guidance mechanism to adaptively select features and reduce communication overhead. Due to the absence of suitable benchmark datasets, we extend three datasets for evaluation: two virtual datasets (Air-to-Pred-Occ and UAV3D-Occ) and one real-world dataset (GauUScene-Occ). Experiments results demonstrate that our method achieves state-of-the-art accuracy, significantly outperforming existing collaborative methods while reducing communication overhead to only a fraction of previous approaches.

cs.CV↗
arXiv2025

Protein as a Second Language for LLMs

Deciphering the function of unseen protein sequences is a fundamental challenge with broad scientific impact, yet most existing methods depend on task-specific adapters or large-scale supervised fine-tuning. We introduce the "Protein-as-Second-Language" framework, which reformulates amino-acid sequences as sentences in a novel symbolic language that large language models can interpret through contextual exemplars. Our approach adaptively constructs sequence-question-answer triples that reveal functional cues in a zero-shot setting, without any further training. To support this process, we curate a bilingual corpus of 79,926 protein-QA instances spanning attribute prediction, descriptive understanding, and extended reasoning. Empirically, our method delivers consistent gains across diverse open-source LLMs and GPT-4, achieving up to 17.2% ROUGE-L improvement (average +7%) and even surpassing fine-tuned protein-specific language models. These results highlight that generic LLMs, when guided with protein-as-language cues, can outperform domain-specialized models, offering a scalable pathway for protein understanding in foundation models.

cs.LG↗