arXiv subjects
Publications and source records attributed to Zhao Zhang.
For a interval family $\calI$, a point set $X\subseteq\mathbb R$, demands $d:X\to\Z_{\ge0}$, and $p\in\Z_{\ge0}$, partial interval multicover asks for a subfamily $S\subseteq\calI$ minimizing $|S|$ subject to $$ |\{x\in X:\dep_S(x)\ge d(x)\}|\ge p, \qquad \dep_S(x)=|\{I\in S:x\in I\}|. $$ We prove that the unweighted decision problem is NP-complete even for unit intervals, answering the open question in~\cite{Li2026}.
We investigate how a two-dimensional spatial embedding changes the entanglement observed across geometrically local bipartitions of a one-dimensional quantum spin chain. A $q$-deformed colored Motzkin chain is placed along a spiral-inspired Hamiltonian path that visits $N(L)=Θ(L^2)$ sites of a square-lattice region while retaining a path interaction graph. Straight ambient-space cuts then correspond to unions of $Θ(L)$ disconnected intervals in the chain ordering. We derive the scaling of a height-variation proxy for the leading color-pairing contribution and predict ambient-space entropy scaling proportional to $L^2$, $L$, and $O(1)$ for $q>1$, $q=1$, and $q<1$, respectively. We evaluate this proxy numerically using the matrix-product-state representation and verify its predicted finite-size scaling. The embedded and original Hamiltonians have exactly identical spectra and spectral gaps at fixed chain length. We also describe a sparse tensor-network representation and propose a two-chain junction extension. These results isolate the roles of interaction connectivity, spatial embedding, and subsystem geometry in entanglement scaling.
Integrating infinitesimal generators of local dynamics into finite-time evolution is generally harder than extracting a local Hamiltonian from a quantum gate. For integrable one-dimensional quantum many-body systems, a recent bootstrap constructs one-parameter $R$-operators directly from Hamiltonians satisfying the Reshetikhin condition. However, this condition excludes important models such as the Hubbard model, whose two-parameter $R$-operator was obtained by Shastry from a decorated Yang--Baxter equation obeyed by its free-fermion sector. Here, I extend the bootstrap to $R$-operators constrained by both the ordinary and decorated Yang--Baxter equations. Their joint expansion produces additional Hamiltonian-level necessary conditions and recursively determines candidate higher-order terms, subject to two-site locality at each order. I also derive compatibility and factorization conditions under which a Shastry-type superposition solves the $RLL$ relation after deformation by an onsite involution. The construction recovers the known two-parameter $R$-operators of the Hubbard and XYh models, whose full Yang--Baxter equations follow from established tetrahedral Zamolodchikov algebra relations. For a superconducting Hubbard deformation, the scalar equations are generically incompatible, placing it outside this framework. These results provide diagnostics and a systematic route to fermionic $R$-operators beyond the Reshetikhin condition. General sufficiency, convergence of the formal bootstrap, and sufficiency of the deformation conditions remain open.
Backpropagation remains the dominant training paradigm, using backward gradients and local updates that can suffer from vanishing or exploding gradients in deep networks. Although the Forward-Forward algorithm avoids conventional backpropagation, it remains mainly local or layer-wise, motivating global whole-network learning using forward evaluations only. For shared-kernel wavelet neural networks, we therefore propose a CMA-ES-based global forward-only learning framework. We further introduce structure-aware parameter decomposition and channel-specific credit assignment, factorizing the search into three subspaces. On BSDS500, our method reduces test MSE from 4.1783 to 4.0130 and improves PSNR from 42.621 to 42.787~dB over Adam/BP without backward propagation. Beyond these gains, training requires only 6 training-set coverages, compared with 1500 epochs for BP. Ablations show that structural decomposition and aligned credit assignment are more important than increasing population size, demonstrating the feasibility of whole-network forward-only learning for compact reconstruction networks. These results suggest a potential route toward forward-only optimization of compact reconstruction models for resource-constrained imaging and edge or near-sensor systems.
The Minimum Shared-Power Edge Cut (MSPEC) problem seeks to disconnect two designated terminals with minimum total power by assigning powers to graph vertices and deleting every edge for which the sum of the powers at its endpoints reaches its edge weight. Previous work gave a fully polynomial-time approximation scheme for this problem, while its exact computational complexity remained open. We prove that general MSPEC with nonnegative edge weights can be solved exactly in polynomial time. Specifically, we first use linear programming duality for each fixed cut to reformulate MSPEC as minimizing the maximum terminal-relaxed matching weight over all $s$-$t$ cuts. Then, we construct a terminal-copy extended graph that transforms the maximum-weight matching model with unconstrained terminal capacities for a fixed cut into an equivalent ordinary maximum-weight matching model. Next, we extend the submodularity of the unweighted maximum-matching cut function to nonnegative edge weights and prove that the restriction obtained by fixing the terminal copies on prescribed sides remains submodular. Consequently, MSPEC reduces to exact submodular function minimization. Once an optimal cut is obtained, a corresponding optimal power assignment can be recovered. This result resolves the open question concerning the exact computational complexity of general MSPEC.
Multi-drone collaborative detection aims to enhance target recognition by leveraging complementary observations from overlapping aerial viewpoints. Existing approaches typically follow a two-stage pipeline that first establishes cross-view correspondences and then performs fusion-based recognition. However, this decomposition suffers from inconsistent optimization objectives, redundant pairwise modeling, and irreversible error propagation, where incorrect associations inevitably undermine subsequent recognition performance. To overcome these limitations, we propose the Coupled Association-Recognition Network (CARNet), a unified framework that jointly optimizes cross-view target association and recognition fusion through a shared target-pair representation. To improve association reliability in densely populated scenes with visually similar targets, CARNet explicitly incorporates spatial cues into pairwise relation modeling. Moreover, we introduce two complementary causal-inspired intervention strategies: the first enhances cross-view association by emphasizing appearance-confusing pairs, while the second bolsters the robustness of fusion-based recognition by learning stable multi-view representations under incomplete visual evidence. In addition, we present MDMT-CD, a fine-grained extension of the MDMT dataset that enriches the original vehicle annotations with detailed category labels, providing the first benchmark specifically designed for multi-drone collaborative vehicle detection. Extensive experiments demonstrate that CARNet consistently outperforms existing state-of-the-art methods. The MDMT-CD dataset and source code will be available at https://github.com/bigdreaming11/CARNet.
Providing more coherent light with space-time coupling offers unprecedented opportunities across diverse scientific disciplines and areas of technology. However, in many applications, tailored laser beams with partial coherence in space and time are desired. Here, we demonstrate a programmable approach for simultaneously shaping the intensity and coherence distributions of spatiotemporally coupled partially coherent wave packets. By jointly engineering the fractal speckle of a dynamic scatterer and applying tailored modulation of the illumination function, we experimentally generate partially coherent laser beams with arbitrarily designed intensity and nontrivial coherence distributions. The proposed approach is compatible with ultrafast laser pulses, enabling the on-demand generation of partially coherent space-time wave packets. The resulting spatiotemporally partially coherent wave packets exhibit non-separable space-time intensity profiles together with engineered spatiotemporal correlation distributions. This work provides a versatile framework for tailoring partially coherent light in space and time, with varied applications in relativistic laser-matter interaction, passive imaging and optical manipulation.
Human motion defines an action, while a camera trajectory determines how it is presented. Camera generation for a given human motion and joint human-camera generation are usually treated as separate tasks, although both share an asymmetric dependency: human motion can be generated independently, whereas the camera responds to the realized action. We introduce AESOP, a unified framework with an independent human pathway and a shared human-conditioned camera generator. Its asymmetric architecture serves both tasks while preserving the human output during camera generation. Although human context anchors the shot to the action and camera text describes its movement, translation intensity remains underspecified. We therefore construct trajectory pairs that differ in camera translation magnitude while sharing human motion and camera text, then use these pairs to learn an explicit intensity condition. Experiments on the PulpMotion dataset demonstrate strong camera distributional and framing quality in both tasks and effective control over camera translation intensity. Project page is available at https://ripemangobox.github.io/AESOP.
Surrogate modelling is important to enhance the computational efficiency of uncertainty quantification for physical processes involving random parameters. The advantage of Fourier neural operator (FNO) compared to conventional surrogates is its cross-resolution prediction capability. In the current study, a cross-resolution surrogate based on FNO is built for two-phase transient Darcy flows in porous media with random parameters. A multiresolution Sobolev learning method is proposed for FNO to enhance the cross-resolution prediction accuracy by incorporating first and second-order gradients in the loss function using training samples of different resolutions. The gradients are approximated by resolution-related finite difference schemes compatible with the grid for numerical simulation. Experimental results demonstrate the effectiveness of the new method in training FNO for building cross-resolution surrogates.
Gradient communication is a primary scaling bottleneck in large language model (LLM) pretraining. Communicating gradients in low-precision formats, such as FP8 and NVFP4, can significantly reduce the communication volume. Existing methods quantize gradients via linear or nonlinear mappings in Euclidean space, often degrading model performance because highly anisotropic gradients incur direction-dependent distortion. We present GeoFP8, a geometry-informed gradient scaling method that performs low-precision communication in geometry-aware coordinates. By transforming gradients into a near-isotropic space before quantization, GeoFP8 makes low-precision representations substantially more faithful to their high-precision counterparts. GeoFP8 only changes the coordinate system used for low-precision gradient communication and does not change the optimizer, training recipe, communication collective, or low-precision format. We also develop a simplified geometry-aware transformation algorithm with low-rank approximation and selective application to balance the computation overhead and communication reduction. We examine the empirical convergence of GeoFP8 using Llama-300M and Llama-600M models. Our results show that GeoFP8 reduces the end-to-end pretraining time of Llama-600M by 7.6% on 64 NVIDIA GH200 Superchips, while improving the downstream task preservation profile over direct Euclidean FP8 communication under the same optimizer and communication path.
We report an improved absolute frequency measurement of the $^{176}$Lu$^+\,(^{3}\mathrm{D}_1)$ optical frequency standard, evaluated via a remote link to the NRC-FCs2 caesium fountain primary frequency standard. Operating a single ion clock with 94.2% uptime over 10 days, and using an ambiguity-resolved precise point positioning (PPP-AR) link over the Global Positioning System, we determine an absolute frequency of $353\,638\,794\,073\,800.332(91)\,$Hz at a fractional uncertainty of $2.6 \times 10^{-16}$. This agrees with our previous result, which underpins the Comité international des poids et mesures recommended frequency value, and reduces the uncertainty by a factor of 3.6.
We report precision measurement of the unperturbed ${^{3}}D_1$ microwave transition frequencies in $^{176}\mathrm{Lu}^+$ to a fractional uncertainty of $4\times10^{-14}$. We find the $|F,m_F\rangle=|8,0\rangle$ to $|7,0\rangle$ hyperfine transition frequency to be $10\,491\,519\,945.228\,82(38)\,$Hz and the $|7,0\rangle$ to $|6,0\rangle$ transition frequency to be $11\,290\,004\,289.881\,61(36)\,$ Hz. At this precision we are able to observe the hyperfine-mediated effects in the ratio of the quadrupole shifts, from which we can directly infer the residual quadrupole moment after $^3D_1$ hyperfine averaging. We find a residual quadrupole moment of ${-2.48(23)\times10^{-4}}\,e a_0^2$, consistent with a previous assessment using a different and less direct method. With the unperturbed microwave frequencies accurately known, the residual quadrupole shift for a $^{176}\mathrm{Lu}^+$ ($^3D_1$) optical frequency standard can henceforth be readily evaluated to $<10^{-20}$ uncertainty by routine ${^{3}}{D}_1$ microwave spectroscopy.
Large-model parameters have grown beyond the capacity of a single xPU, dispersing across multiple xPUs spanning distinct host domains, where xPU-to-xPU communication dominates overall system efficiency. Existing scale-up interconnect remains inadequate: network-based solutions built on Ethernet--such as RoCE (RDMA over Converged Ethernet)--introduce specific message-semantics and protocol-stack characteristics, and rely on fragmented per-host addressing, while conventional host-based fabrics are confined to a single host domain and lack cross-host unified addressing. This paper presents Co-Fabric, a bus-based interconnect that, unlike conventional bus designs, breaks host-domain boundaries to deliver unified xPU interconnection for scale-up superpods. Co-Fabric makes three contributions: a streamlined four-layer protocol stack achieving nanosecond-scale processing latency with native reliability; a cross-domain scaling and P2P mechanism that routes using port identifiers embedded in the packet header; and a unified address space built on shadow-device auto-enumeration. On a 64-xPU 3D-Mesh system, Co-Fabric cuts inter-node communication latency by over 50% and improves bandwidth by 2-5x over RoCE, accelerating DeepSeek R1 inference by 30%-80%. Moreover, since its streamlined four-layer protocol stack and higher data-communication efficiency reduce protocol and processing overhead relative to the Ethernet-based RoCE stack, Co-Fabric cuts the cost and power of the interconnect itself by up to 80% and 5%, respectively. These results demonstrate Co-Fabric's advantage for AI computing centers.
Image-to-layer decomposition converts a flattened image into editable RGBA layers, enabling element-level editing in design workflows. Existing diffusion-based systems typically adapt large pretrained text-to-image (T2I) models and introduce RGBA autoencoders or variable-layer architectural modules. We revisit this design choice and ask whether layer decomposition truly requires these heavyweight components. We introduce PixelART, a pixel-space rectified-flow Transformer trained from scratch for image-to-layer (I2L) decomposition. PixelART directly denoises regional RGBA pixel patches using a single-stream multi-modal diffusion Transformer, avoiding RGBA-VAEs, pretrained T2I backbones, and layer-specific decoders. We identify a key property of the task: high-noise timesteps determine layer assignment and coarse layer organization, while low-noise timesteps mainly refine color, alpha, texture, and boundaries. Based on this observation, we propose a terminal-boosted timestep sampling strategy to increase training coverage in the high-noise layer assignment regime. Trained on 4M multi-layer design templates, PixelART achieves state-of-the-art layer decomposition and composite reconstruction on Design-Multi-Layer-Bench and LICA with over 80% fewer parameters, 98% lower latency, and 85% lower memory than the recent Qwen-Image-Layered model. Ablation experiments show that pixel-space $\mathbf{x}$-prediction, terminal-boosted timestep sampling, and data/model scaling are critical, while T2I pretraining brings marginal benefits to the I2L task.
Click-through rate (CTR) prediction serves as a cornerstone of recommender systems. Despite the strong performance of current CTR models based on user behavior modeling, they are still severely limited by interaction sparsity, especially in low-active user scenarios. To address this issue, data augmentation of user behavior is a promising research direction. However, existing data augmentation methods heavily rely on collaborative signals while overlooking the rich multimodal features of items, leading to insufficient modeling of low-active users. To alleviate this problem, we propose a novel framework \textbf{MARS} (\textbf{M}odality-\textbf{A}ligned \textbf{R}etrieval for \textbf{S}equence Augmented CTR Prediction). MARS utilizes a Stein kernel-based approach to align text and image features into a unified and unbiased semantic space to construct multimodal user embeddings. Subsequently, each low-active user's behavior sequence is augmented by retrieving, filtering, and concentrating the most similar behavior sequence of high-active users via multimodal user embeddings. Validated by extensive offline experiments and online A/B tests, our framework MARS consistently outperforms state-of-the-art baselines and achieves substantial growth on core business metrics within Kuaishou~\footnote{https://www.kuaishou.com/}. Consequently, MARS has been successfully deployed, serving the main traffic for hundreds of millions of users. To ensure reproducibility, we provide anonymous access to the implementation code~\footnote{https://github.com/wangshukuan/MARS}.
Compact summaries are a key tool for approximate query processing over large datasets. For range-query workloads, an $\varepsilon$-net provides a small summary that hits every sufficiently large range. However, classical $\varepsilon$-nets only guarantee range validity and do not control the group composition of the selected tuples. As a result, the summary may be range-valid but poorly representative, which can propagate imbalance to downstream query results. Motivated by recent work on fair $\varepsilon$-nets and fair geometric hitting sets \cite{dehghankar2025fair}, we study fairness-aware range summaries under prescribed target group ratios. Different from previous sample-and-repair approach, we propose a target-stratified sampling method. For demographic parity (in which the ratio of fairness is determined by group proportion), our sample size is $O(A_{\varepsilon})$, coinciding with the standard $\varepsilon$-net bound, improving previous bound of $O\!\left(A_\varepsilon\log\frac{k}φ\right)$. For custom-ratio targets (in which the ratio of fairness is determined by manually defined proportion), our sample size is $O(A_Γ)$, where $Γ$ is a parameter measuring the gap between the customized ratio and the demographic parity; we prove that this dependence on $Γ$ is unavoidable, with a worst-case lower bound of $Ω(Γ/\varepsilon)$. Using our target-stratified sampling method, we could improve the previous approximation ratio for the fair geometric hitting set problem by a logarithmic factor, and making use of this result, we could in turn improve the size of custom-ratio fair $\varepsilon$-net. Experiments on real and synthetic datasets demonstrate that our method constructs smaller fair summaries than existing approaches, scales to large datasets and fine-grained group constraints, and improves downstream range query processing.
Wild test-time adaptation (WTTA) updates a source model online under small test batches, concurrent distribution shifts, and time-varying class imbalance. Most WTTA methods derive their adaptation signals, including predictive uncertainty, sample reliability, and local feature geometry, from the model being adapted. When the source model is unreliable under shift, these signals can reinforce its own errors, forming a self-referential loop. We introduce MASA (Multimodal-LLM-Anchored Semantic Adaptation), which complements model-internal evidence with structured semantic descriptions from a frozen multimodal large language model (MLLM). To limit inference cost, MASA queries the MLLM only for a small set of diverse, reliability-ranked anchors. The resulting descriptions capture the object family and nuisance factors such as style, viewpoint, and occlusion. MASA encodes these descriptions, propagates them to neighboring test samples, and stores the resulting visual-semantic information in an online prototype memory. Descriptor-aware retrieval from this memory provides an auxiliary target for lightweight adaptation of normalization-affine parameters. We evaluate MASA on the WTTA ImageNet-C benchmark under limited-batch, mixed-domain, and imbalanced-label-shift settings with ResNet and ViT backbones.
Retrieval-augmented scientific synthesis aims to answer complex research questions by integrating information from multiple papers into comprehensive and well-grounded responses. Since the generator can only synthesize the information selected and organized by the reranker, the quality of the generated synthesis depends critically on the reranked results. However, most rerankers operate at the passage level, which leaves key methodological, empirical, and comparative information buried in long and flat contexts, weakening the grounding of generated claims. Moreover, existing rerankers mainly rely on independent query-candidate scoring which overlooks complementary, contextual, and contrasting relations across scientific candidates, limiting information coverage and the comprehensiveness of the resulting synthesis. To address these limitations, we propose LoG-Reranker, a local-to-global sentence-level graph reranking framework for scientific synthesis. LoG-Reranker performs role-aware local scoring to identify fine-grained, query-relevant sentences and then models their relations on a sentence graph across the candidate set to globally refine sentence rankings. Top-ranked sentences and their connected neighbors are organized into a structured input context for generator to produce more grounded and comprehensive synthesis. Extensive experiments on scientific synthesis and reranking benchmarks show that LoG-Reranker consistently outperforms competitive rerankers, yielding more reliable rankings and improving the quality of generated synthesis.