Searcharxiv⌕ Search

arXiv subjects

Xinbing Wang

Publications and source records attributed to Xinbing Wang.

At least 19 recordsLinked to original sources
arXiv2026

HyperIntent: Tracking Intent Shifts for Efficient Reasoning on Dynamic Textual Graphs

Dynamic web interactions continuously evolve as new edges arrive, offering large language models (LLMs) great potential for reasoning over edge-level temporal semantics. However, continuously arriving edges are often repetitive, obscuring the key structures that truly reflect temporal changes and limiting the efficiency of LLMs. A key insight is that beneath the dense stream of edge arrivals lies a sparse evolution of temporal intent, enabling LLMs to be responsively triggered only when such intent shifts, while otherwise inheriting the previous state. To this end, we propose a hypergraph-based framework named HyperIntent, which enhances dynamic graph reasoning by parsing differentiated intents, quantifying their temporal discrepancies, and performing shift-triggered semantic updating. Experimental results on destination node retrieval and future link prediction demonstrate that HyperIntent consistently achieves the best overall performance, with average relative Hits@1 gains of 34.0% and 25.4% over vanilla TGAT and DyGFormer, respectively. Meanwhile, HyperIntent reduces LLM token consumption by 54.4%-95.9% across six datasets while achieving leading predictive performance. In-depth studies further confirm the necessity of shift triggering for identifying structural and semantic temporal changes. The code will be made publicly available upon completion of the review process.

cs.GR↗
arXiv2026

Distill the Visual Evidence, Not Just the Answer: Cross-World On-Policy Distillation for Vision-Language Models

A central goal of vision-language model (VLM) distillation is to transfer both the teacher's language capabilities and its visual understanding. However, existing methods primarily supervise the student's output, leaving visual understanding implicit. Our analysis reveals that a student can match the teacher's answer without relying on the same visual evidence, raising the question: how can we ensure the student responds to the visual information that actually determines the answer? To this end, we propose \textbf{Cross-World On-Policy Distillation (CW-OPD)}, which explicitly supervises the student's response to changes in visual evidence. For each example, CW-OPD constructs two visual worlds that share the question and scene context but differ in answer-critical evidence, yielding different answers. We perform on-policy distillation in both worlds and distill the teacher's cross-world belief transition, encouraging the student to match not only \emph{what} the teacher predicts but also \emph{why} its prediction changes with the evidence. A gradient analysis shows that this term is invariant to errors shared by both worlds and supplies a corrective signal invisible to endpoint matching alone. In this way, CW-OPD makes reliance on the relevant visual evidence an explicit distillation target rather than an implicit consequence of output matching. To diagnose whether a model truly grounds its answers in visual evidence, we introduce CWBench, which measures cross-world consistency via Cross-World Pair Accuracy (CWPA). Experiments on Qwen3.5-4B show that CW-OPD outperforms the strongest baseline by \textbf{1.2} points on average, and the 4B student exceeds DeepSeek-V4.1 (552B) by \textbf{22.4} CWPA points on CWBench. Code is released in https://github.com/baokou-fw2/CWAD.

cs.CV↗
arXiv2026

LEAP: Unlocking dLLM Parallelism via Lookahead Early-Convergence Token Detection

Diffusion Language Models (dLLMs) have garnered significant attention for their potential in highly parallel processing. The parallel capabilities of existing dLLMs stem from the assumption of conditional independence at high confidence levels, which ensures negligible discrepancy between the marginal and joint distributions. However, the stringent confidence thresholds required to preserve accuracy severely constrain the scalability of parallelism. Through systematic token-level statistical analysis, we reveal that a substantial proportion of tokens converge to their correct predictions early in the denoising process yet fail to reach standard confidence thresholds, confirming that current confidence-based criteria are overly conservative. In response, we introduce LEAP (Lookahead Early-Convergence Token Detection for Accelerated Parallel Decoding). LEAP is a training-free, plug-and-play method that leverages future context filtering and multi-sequence superposition to detect early-converging tokens. By validating the alignment between early convergence and correctness, we enable reliable early decoding of these tokens. Benchmarking across diverse domains demonstrates that LEAP significantly lowers inference latency and decoding steps. Compared to confidence-based decoding, the average number of denoising steps is reduced by about 30%. On the GSM8K dataset, combining LEAP with dParallel accelerates decoding to 7.2 tokens per step while preserving model precision. LEAP effectively breaks the reliance on high-confidence priors, offering a novel paradigm for parallel decoding.

cs.LG↗
arXiv2026

On the Design of Physical Topology in OCS-based Clusters

Optical Circuit Switches (OCSes) are increasingly deployed in data center networks and AI clusters. By wiring the electrical ports of GPUs and electrical switches to the OCS ports, the physical interconnection pattern defines a \emph{physical topology}. \textbf{Topology Engineering (ToE)} seeks an OCS configuration that optimizes a traffic objective given traffic demands and physical constraints, but solving this joint problem is computationally expensive. A common approach is to \emph{decouple} the problem into two steps: first, compute a \emph{logical topology} that satisfies the traffic demands without considering the physical topology constraints and then map it onto the physical topology to obtain the OCS configuration. This, however, introduces a quality loss issue: the resulting logical topology sometimes cannot be realized exactly under the given physical topology. To address this problem, we show that it stems from a property missing in the physical topology. We propose \textbf{Decoupling Optimality} and the necessary and sufficient conditions under which a decoupled method will not lose quality. Examining existing physical topologies against these conditions reveals that they either fail them, or satisfy them at the cost of cluster scale or deployment generality. Our proposed \textbf{Cross Wiring} meets all the conditions, without sacrificing cluster scale or adding devices. On a 128-NPU testbed, the lack of \emph{Decoupling Optimality} increases training iteration time by up to 39.5\%, and a microbenchmark validates the importance of \emph{Decoupling Optimality} in both solution quality and solving speed. A testbed built from two MEMS OCSes validates the engineering feasibility of Cross Wiring.

cs.NI↗
arXiv2026

DoPR: Reusable Compressed Document Prefixes for Efficient LLM Reranking

Large language models (LLMs) are effective rerankers, but pointwise reranking repeatedly processes the same document across different queries, causing substantial redundant document-side computation. We propose \textbf{DoPR}, a compressed document prefix framework that decouples offline document processing from online reranking. DoPR first selects query-independent document representations and converts them into compressed document prefix states, which are precomputed offline and reused whenever the document is retrieved. During online reranking, the model scores each query-document pair by processing only the query and scoring token, with document information supplied by the stored prefix states. This design reduces online cost through both document-side compression and cross-query prefix-state reuse. Experiments on TREC DL, BEIR, and BRIGHT with Qwen3 models from $0.6$B to $8$B show that DoPR achieves up to 8.0$\times$ online document-side memory reduction and up to 8.04$\times$ latency speedup, while retaining \textbf{97.1\%-99.5\%} of the average NDCG@10 of matched full-document rerankers.

cs.IR↗
arXiv2026

HAP: Head-Adaptive Visual Token Pruning via Cross-Modal Alignment

Recent Vision-Language Models encode high-resolution images into long visual token sequences, incurring prohibitive prefill costs. To compress them, existing methods score each visual token by averaging text-to-visual attention uniformly across all heads, which assumes every head matches the query. However, our empirical analysis shows that misaligned heads dominate the average, amplifying background tokens and drowning out fine-grained cues. To address this, we propose PAQ (Prompt-Grounded Attention Quality), a metric quantifying how well each head aligns the prompt with image regions. Built on PAQ, our pruning proceeds in three stages. Given a target FLOPs budget, we first partition the transformer layers into groups and allocate a visual token budget to each. Within each group, we then aggregate per-head attention maps via PAQ-weighted softmax into a group-level matrix. Finally, we score visual tokens by this matrix's magnitude and retain the allocated budget per group. By weighting heads with PAQ, our method scores tokens by attention signals that more faithfully reflect prompt relevance, rather than diluting them through uniform averaging. Across 18 benchmarks, our method delivers state-of-the-art trade-offs. Specifically, on LLaVA-1.5-7B (9 tasks), retaining only \textbf{5.6\%} tokens preserves \textbf{99.1\%} of the original performance, surpassing the strongest baseline AutoPrune by 4.2 points. Code is available in https://github.com/baokou-fw2/HAP.

cs.CV↗
arXiv2026

ENCORE: Entropy-Guided Cropping and Attention Regularization for Robust Vision--Language Understanding

Vision-Language Models (VLMs) perform well on diverse vision-language tasks, but transformer-based visual encoders split images into fixed-resolution sub-images, compromising object integrity in lightweight VLMs. Existing methods only focus on the visual modality and fail to dynamically preserve the integrity of prompt-relevant regions, limiting performance. In this work, we observe that the early-layer image-text entropy of cross-modal attention strongly correlates with answer grounding quality and task accuracy. Building on this finding, we propose \textbf{ENCORE}, an entropy-guided framework with two components: At inference, an \textbf{Entropy-based Cropping Strategy} (ECS) evaluates a small set of candidate crops and selects the one with minimal entropy, preserving contiguous regions relevant to the prompt. At training, \textbf{Entropy Regularization Training} (ERT) augments next-token prediction with an entropy term that sharpens attention on key visual tokens while down-weighting irrelevant ones. Experiments on ten VQA benchmarks show that ENCORE, fine-tuning only 0.14\% of parameters, achieves an average 1.43\% accuracy gain and state-of-the-art performance among recent 2B-parameter VLMs. Our code is released in https://github.com/baokou-fw2/ENCORE.

cs.CV↗
arXiv2026

GRAND-HC: Graph-Refined Author Name Disambiguation

From-Scratch Name Disambiguation (SND) groups papers sharing an ambiguous name into clusters of distinct real-world authors. Existing methods suffer from two critical limitations: (1) inherent long-tailed author distribution biases representation learning, causing over-merging of tail authors; (2) existing cluster number estimation methods are unreliable for long paper sequences, hindering large-scale deployment. We propose \textbf{GRAND-HC}, a complete end-to-end SND framework. We construct a heterogeneous paper graph via co-author, co-organization, and co-venue relations, using a graph attention network as the embedding backbone. \textbf{Harmony Contrastive Learning (HCL)} dynamically reweights training loss to suppress overfitting to prolific authors, learning discriminative embeddings. A \textbf{Graph-Refined Distance Matrix (GRDM)} leverages graph topology to optimize pairwise distances, further preventing tail author over-merging. Meanwhile, a lightweight \textbf{Paper Compression Module (PCM)} achieves accurate cluster number estimation across varying scales. Finally, Hierarchical Agglomerative Clustering outputs the final clusters. Extensive experiments demonstrate state-of-the-art macro F1 performance. GRAND-HC has been deployed in a billion-scale academic database. Source code: https://github.com/baokou-fw2/GRAND-HC.

cs.IR↗
arXiv2026

Information Source Detection with Limited Time Knowledge

This paper investigates the problem of utilizing network topology and partial timestamps to detect the information source in a network. The problem incurs prohibitive cost under canonical maximum likelihood estimation (MLE) of the source due to the exponential number of possible infection paths. Our main idea of source detection, however, is to approximate the MLE by an alternative infection path based estimator, the essence of which is to identify the most likely infection path that is consistent with observed timestamps. The source node associated with that infection path is viewed as the estimated source $\hat{v}$. We first study the case of tree topology, where by transforming the infection path based estimator into a linear integer programming, we find a reduced search region that remarkably improves the time efficiency. Within this reduced search region, the estimator $\hat{v}$ is provably always on a path which we term as \emph{candidate path}. This notion enables us to analyze the distribution of $d(v^{\ast},\hat{v})$, the error distance between $\hat{v}$ and the true source $v^{\ast}$, on arbitrary tree, which allows us to obtain for the first time, in the literature provable performance guarantee of the estimator under limited timestamps. Specifically, on the infinite $g$-regular tree with uniform sampled timestamps, we get a refined performance guarantee in the sense of a constant bounded $d(v^{\ast},\hat{v})$. By virtue of time labeled BFS tree, the estimator still performs fairly well when extended to more general graphs. Experiments on both synthetic and real datasets further demonstrate the superior performance of our proposed algorithms.

cs.SI↗
arXiv2026

Improving LLM Reasoning with Homophily-aware Structural and Semantic Text-Attributed Graph Compression

Large language models (LLMs) have demonstrated promising capabilities in Text-Attributed Graph (TAG) understanding. Recent studies typically focus on verbalizing the graph structures via handcrafted prompts, feeding the target node and its neighborhood context into LLMs. However, constrained by the context window, existing methods mainly resort to random sampling, often implemented via dropping node/edge randomly, which inevitably introduces noise and cause reasoning instability. We argue that graphs inherently contain rich structural and semantic information, and that their effective exploitation can unlock potential gains in LLMs reasoning performance. To this end, we propose Homophily-aware Structural and Semantic Compression for LLMs (HS2C), a framework centered on exploiting graph homophily. Structurally, guided by the principle of Structural Entropy minimization, we perform a global hierarchical partition that decodes the graph's essential topology. This partition identifies naturally cohesive, homophilic communities, while discarding stochastic connectivity noise. Semantically, we deliver the detected structural homophily to the LLM, empowering it to perform differentiated semantic aggregation based on predefined community type. This process compresses redundant background contexts into concise community-level consensus, selectively preserving semantically homophilic information aligned with the target nodes. Extensive experiments on 10 node-level benchmarks across LLMs of varying sizes and families demonstrate that, by feeding LLMs with structurally and semantically compressed inputs, HS2C simultaneously enhances the compression rate and downstream inference accuracy, validating its superiority and scalability. Extensions to 7 diverse graph-level benchmarks further consolidate HS2C's task generalizability.

cs.AI↗
arXiv2026

Depth-Attention: Cross-Layer Value Mixing for Language Models

Self-attention selects information freely across the sequence, but across depth, Transformers merely add each layer's output to the residual stream, so later layers cannot selectively reuse earlier-layer representations. Recent cross-layer methods improve this flow but operate on hidden states outside attention, adding state beyond the key-value cache at inference--a cost that becomes increasingly salient as modern LLMs compress the cache with grouped-query and multi-head latent attention. We introduce Depth-Attention, which performs this selection inside the attention module itself: before a layer attends over the sequence, its query attends over the keys of earlier layers at the same token position and mixes their values into the value that self-attention then reads. Because Depth-Attention reuses the standard attention queries, keys, and value-cache slots, storing depth-mixed values in place of the original values, it adds no parameters and introduces no persistent inference state beyond the standard key-value cache--the same cache size as a vanilla decoder and less than hidden-state-based cross-layer methods. On Qwen3-style decoders at 1.5B and 3B parameters, Depth-Attention attains the lowest perplexity and the highest average downstream accuracy, improving over the vanilla Transformer by up to 2.3 accuracy points and surpassing strong cross-layer baselines in perplexity and average accuracy, while adding under 0.01% extra arithmetic FLOPs and no additional persistent inference state. The gains hold from 360M to 3B parameters and extend to looped Transformers.

cs.CL↗
arXiv2026

Compass: Navigating Global Marine Lead Data Integration through Expert-Guided LLM Agent

Marine lead (Pb) and its isotopes are critical tracers for ocean circulation and anthropogenic pollution, yet in-situ observations remain costly and sparse. While vast historical records exist, they lie buried within the unstructured content of academic papers, creating "data silos" inaccessible to comprehensive analysis. Manual extraction is unscalable, while general-purpose Large Language Models (LLMs) lack the necessary domain-specific knowledge, leading to hallucinations and scientifically invalid outputs. To address this, we introduce an expert-guided adaptation approach that enables LLMs to perform rigorous scientific data extraction without fine-tuning. We operationalize this approach through Compass, an LLM agent framework enhanced by a Knowledge Tree co-designed with marine scientists, which decomposes complex tasks into verifiable steps, guiding the agent's reasoning to ensure scientific validity. Deploying Compass across a corpus of over 230,000 relevant open-access papers, we successfully extract 3,751 previously unincorporated Pb records. This effort establishes the largest integrated marine Pb database to date. Beyond standard metrics, Compass demonstrates superior reliability through multi-layered validation, achieving 92% accuracy as confirmed through expert manual verification. The newly integrated data expand coverage in previously under-sampled regions such as the East China Sea and the Southern Ocean, providing an enriched data foundation for future scientific discoveries. We release an interactive visualization platform to facilitate open scientific access. Our work demonstrates that expert-guided agents can effectively bridge the gap between general-purpose LLMs and high-stakes scientific domains, enabling scalable data discovery in geosciences.

cs.AI↗
arXiv2026

CRoFT: Robust Fine-Tuning with Concurrent Optimization for OOD Generalization and Open-Set OOD Detection

Recent vision-language pre-trained models (VL-PTMs) have shown remarkable success in open-vocabulary tasks. However, downstream use cases often involve further fine-tuning of VL-PTMs, which may distort their general knowledge and impair their ability to handle distribution shifts. In real-world scenarios, machine learning systems inevitably encounter both covariate shifts (e.g., changes in image styles) and semantic shifts (e.g., test-time unseen classes). This highlights the importance of enhancing out-of-distribution (OOD) generalization on covariate shifts and simultaneously detecting semantic-shifted unseen classes. Thus a critical but underexplored question arises: How to improve VL-PTMs' generalization ability to closed-set OOD data, while effectively detecting open-set unseen classes during fine-tuning? In this paper, we propose a novel objective function of OOD detection that also serves to improve OOD generalization. We show that minimizing the gradient magnitude of energy scores on training data leads to domain-consistent Hessians of classification loss, a strong indicator for OOD generalization revealed by theoretical analysis. Based on this finding, we have developed a unified fine-tuning framework that allows for concurrent optimization of both tasks. Extensive experiments have demonstrated the superiority of our method. The code is available at https://github.com/LinLLLL/CRoFT.

cs.CV↗
arXiv2026

$Δ\mathrm{Energy}$: Optimizing Energy Change During Vision-Language Alignment Improves both OOD Detection and OOD Generalization

Recent approaches for vision-language models (VLMs) have shown remarkable success in achieving fast downstream adaptation. When applied to real-world downstream tasks, VLMs inevitably encounter both the in-distribution (ID) data and out-of-distribution (OOD) data. The OOD datasets often include both covariate shifts (e.g., known classes with changes in image styles) and semantic shifts (e.g., test-time unseen classes). This highlights the importance of improving VLMs' generalization ability to covariate-shifted OOD data, while effectively detecting open-set semantic-shifted OOD classes. In this paper, inspired by the substantial energy change observed in closed-set data when re-aligning vision-language modalities (specifically by directly reducing the maximum cosine similarity to a low value), we introduce a novel OOD score, named ΔEnergy. ΔEnergy significantly outperforms the vanilla energy-based OOD score and provides a more reliable approach for OOD detection. Furthermore, ΔEnergy can simultaneously improve OOD generalization under covariate shifts, which is achieved by lower-bound maximization for ΔEnergy (termed EBM). EBM is theoretically proven to not only enhance OOD detection but also yields a domain-consistent Hessian, which serves as a strong indicator for OOD generalization. Based on this finding, we developed a unified fine-tuning framework that allows for improving VLMs' robustness in both OOD generalization and OOD detection. Extensive experiments on challenging OOD detection and generalization benchmarks demonstrate the superiority of our method, outperforming recent approaches by 10% to 25% in AUROC.

cs.CV↗
arXiv2026

SLASH the Sink: Sharpening Structural Attention Inside LLMs

Large Language Models (LLMs) show remarkable semantic understanding but often struggle with structural understanding when processing graph topologies in a serialized format. Existing solutions rely on training external graph-based adapters or fine-tuning, which incur high costs and lost generalizability. In this work, we investigate the internal mechanisms of LLMs and present a critical finding: LLMs spontaneously reconstruct the graph's topology internally, evidenced by a distinct "sawtooth" pattern in their attention maps that structurally aligns with the "token-level adjacency matrix". However, this intrinsic structural understanding is diluted by the attention sink. We theoretically formalize this dilution as a representation bottleneck, stemming from a fundamental conflict: the model's anisotropic bias, essential for language tasks, suppresses the topology-aware local aggregation required for graph reasoning. To address this, we propose a training-free solution, named StructuraL Attention SHarpening (SLASH), which amplifies this internal structural understanding via a plug-and-play attention redistribution. Experiments on pure graph tasks and molecular prediction validate that SLASH delivers significant and consistent performance gains across diverse LLMs.

cs.AI↗
arXiv2026

Rethinking Efficient Graph Coarsening via a Non-Selfishness Principle

Graph coarsening is a graph dimensionality reduction technique that aims to construct a smaller and more tractable graph while preserving the essential structural and semantic properties of the original graph. However, most existing methods rely on pair-wise similarity matching, where each node independently searches for its best partner based on global information. This selfishness matching paradigm incurs substantial computational and memory overhead. To address this problem, we shift to a non-selfishness principle that prioritizes the collective interference of neighborhood in coarsening, and propose an efficient method named NOPE, which achieves linear memory consumption and near-linear computational complexity in the number of nodes. Furthermore, we derive a faster variant NOPE*, which reduces O(δ\dot d) interference evaluation to O(d) based on the local isotropy assumption, and consequently alleviates the computational bottleneck for high-degree nodes. Experimental results show that NOPE* achieves 1.8-10\times speedup over NOPE and surpass almost all baselines with 1-3 orders of magnitude acceleration. Meanwhile, learning on coarsened graphs yields comparable performance to original graphs, and can even show superior performance over LLM-based graph reasoning owing to compact graph information. The code can be available at https://github.com/dazonglian/NOPE-main.

cs.LG↗
arXiv2026

Logit-Attention Divergence: Mitigating Position Bias in Multi-Image Retrieval via Attention-Guided Calibration

Multimodal Large Language Models (MLLMs) have shown strong performance in multi-image cross-modal retrieval, yet suffer from severe position bias, where predictions are dominated by input order rather than semantic relevance. Through empirical analysis, we identify a phenomenon termed Logit-Attention Divergence, in which output logits are heavily biased while internal attention maps remain well-aligned with relevant visual evidence. This observation reveals a fundamental limitation of existing logit-level calibration methods such as PriDe. Based on this insight, we propose a training-free, attention-guided debiasing framework that leverages intrinsic attention signals for instance-level correction at inference time, requiring only a minimal calibration set with negligible computational overhead. Experiments on MS-COCO-based benchmarks show that our method substantially improves permutation invariance and achieves state-of-the-art performance, enhancing accuracy by over 40\% compared to baselines. Code is available at https://github.com/brightXian/LAD.

cs.CV↗
arXiv2026

TINS: Test-time ID-prototype-separated Negative Semantics Learning for OOD Detection

Vision-language models enable OOD detection by comparing image alignment with ID labels and negative semantics. Existing negative-label-based methods mainly rely on static negative labels constructed before inference, limiting their ability to cover diverse and evolving OOD concepts. Although test-time expansion provides a natural solution, naively learning negative semantics from potential OOD samples may introduce hard ID contamination. To address this issue, we propose a \textbf{T}est-time \textbf{I}D-prototype-separated \textbf{N}egative \textbf{S}emantics learning method, termed \textbf{TINS}. TINS learns sample-specific negative text embeddings via image-to-text modality inversion and introduces ID-prototype-separated regularization to keep them separated from ID semantics. To further stabilize negative semantics expansion, TINS employs group-wise aggregation scoring and a buffer update strategy. Extensive experiments across Four-OOD, OpenOOD, Temporal-shift, and Various ID settings show consistent improvements over strong baselines. Notably, on the Four-OOD benchmark with ImageNet-1K as ID, TINS reduces the average FPR95 from 14.04\% to 6.72\%. Our code is available at https://github.com/zxk1212/tins.

cs.CV↗