Search arXiv⌕ Search

arXiv subjects

Xu Chen

Publications and source records attributed to Xu Chen.

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

Text-Video Retrieval via Multi-Dimensional Saliency Assessment and Granularity-Aware Query Decomposition

Text-video retrieval, which aims to bridge visual and textual modalities by learning a joint embedding space, has become a crucial task in multimodal intelligence. Despite extensive efforts to mitigate visual redundancy, previous methods typically rely on a single-aspect criterion to assess visual importance, overlooking the multifaceted spatiotemporal nature of video. In addition, encoding text into a single global embedding to align with videos compresses temporal events and spatial entities into a unified representation space, further aggravating cross-modal misalignment. To address these issues, we propose MMTI, a method that jointly mitigates visual redundancy and enables multi-grained text-video interaction to achieve accurate multi-grained semantic alignment. Specifically, a key feature selection (KFS) mechanism adaptively identifies and aggregates informative frames and patches by jointly evaluating multi-dimensional saliency and learnable importance scores, effectively compacting dense visual features and mitigating visual redundancy. Furthermore, our proposed multi-grained text-video interaction module (TVIM) employs a dynamic gating mechanism to decompose the text query into sentence, frame, and patch queries (SFP), enabling multi-grained text-video alignment. Complementary alignment at different granularities is thereby achieved. Extensive experiments on four standard benchmarks demonstrate that our method outperforms state-of-the-art methods.

cs.IR↗
arXiv2026

Structural Process Supervision for Latent Chain-of-Thought Reasoning

Latent reasoning approaches enhance token-level efficiency and robustness by replacing verbose, explicit chain-of-thought (CoT) tokens with compact continuous-space embeddings. However, existing methods lack efficient process supervision over these embeddings, which often leads to representation collapse and uneven information distribution. % However, supervising these latent embeddings is challenging because a fixed number of latent states must capture information from variable-length reasoning traces. To address this, we propose Prototype-Mediated Process Supervision (PMPS), which introduces learnable reasoning prototypes as semantic anchors to provide structural process-level supervision for latent reasoning. PMPS projects latent embeddings and explicit CoT embeddings into a shared prototype space, achieving many-to-many soft alignment between unequal-length representations through prototype assignment. Meanwhile, we introduce a Progressive Sequential Alignment (PSA) module to further guide training: positional priors initially encourage sequential alignment structure, then gradually relax to permit adaptive matching. Experimental results show that PMPS compresses output token length to under 50% of explicit CoT on GSM8K-Aug. Compared to leading baseline SIM-CoT, our method achieves average accuracy gains of 2.51% across different model families. On GPT-2, accuracy of PMPS even surpasses CoT-SFT. On larger models and a more challenging task, PMPS consistently attains the highest accuracy among all latent reasoning methods with comparable output length.

cs.AI↗
arXiv2026

Relevance-Resolution Transfer via Scale-Decomposable Fractional Diffusion for Multi-Length Cross-Modal Hash Retrieval

Cross-modal hashing enables efficient retrieval by encoding heterogeneous data into compact binary codes. Recent methods exploit fine-grained relations encoded in multi-label training structure, yet none of them constrains how those relations survive as consistent candidate rankings in finite, multi-length Hamming spaces, which we term the relevance resolution bottleneck (RRB). To address the RRB, we propose MultiBit, which transfers relevance resolution from multi-label structure to multi-length Hamming spaces. MultiBit first constructs a scale-decomposable fractional relation teacher from dataset-level label co-occurrence and label specificity, and models dependencies from local to long-range over continuous diffusion scales. It then maps the discretized diffusion scales and their quadrature weights to scale-aware bit subblocks of the maximum-length code, organizes the target code lengths as nested prefixes, and aligns their Hamming candidate rankings with the teacher relations. Experiments on multiple benchmarks demonstrate improved retrieval accuracy. Code is available in the supplementary material.

cs.IR↗
arXiv2026

Adaptive Resource Allocation for Effective and Efficient LLM Social Survey Simulation

Large Language Models (LLMs) enable scalable social survey simulation, yet existing pipelines typically use the same strong general-purpose model and a fixed, often large, respondent history for every respondent-question request. This uniform approach overlooks three factors. First, stronger models may rely on their own knowledge rather than respondent-specific evidence while costing more. Second, additional history can help when evidence is limited, but irrelevant responses may add noise and increase input length. Third, the preferred model and history budget can depend on each other. We propose E2Sim, an adaptive resource allocation framework that jointly selects a model and history budget for each request. Given a respondent persona, ranked response history, and target question, a lightweight policy predicts the accuracy and cost of each configuration and selects the most suitable one. We use respondent-history drop-and-swap augmentation to improve robustness to incomplete or variable histories, and a margin-based curriculum that progresses from easier allocation decisions to harder ones. Experiments on four real-world social survey datasets, multiple model pools, and different history-budget spaces show improvements over the oracle best-fixed configurations, with accuracy gains of up to 6.3 percentage points and cost reductions of up to 55.3 percent. Code is available at https://anonymous.4open.science/r/E2Sim-DE66.

cs.CY↗
arXiv2026

SCBO: Semantically Coherent Batching and Ordering for LLM-Based Social Surveys

Large Language Models (LLMs) offer a scalable way to simulate survey respondents using demographic profiles and observed reference responses. However, the conventional approach of predicting one question per prompt repeatedly encodes the same context, limits each target to a narrow set of reference responses, and prevents later predictions from using information in earlier answers. Predicting multiple questions in one prompt can reduce these costs, share a broader pool of references, and let later predictions build on earlier ones. This requires forming coherent batches, selecting shared references, and ordering questions and references effectively. We propose Semantically Coherent Batching and Ordering (SCBO), a training-free framework that addresses these challenges. SCBO first uses an LLM to extract compact semantic representations from survey items and filter out template noise. It then groups related questions into batches and builds a shared reference bank using target-specific retrieval and centroid-based completion. Finally, it orders target questions from easy to hard and arranges references according to their semantic alignment with those questions. Experiments on four large-scale survey datasets and four LLMs show that SCBO substantially reduces token consumption and inference time while generally improving prediction accuracy over a non-batched baseline. Code is available at https://anonymous.4open.science/r/SCBO-41D8.

cs.CL↗
arXiv2026

PACE: A Unified Condense-and-Extract Paradigm for Fast VLM Inference

Vision-Language Models (VLMs) demonstrate exceptional visual reasoning capabilities, yet their inference costs escalate rapidly with the proliferation of visual tokens. Existing visual token pruning methods exhibit two fundamental limitations. First, most approaches operate exclusively post-vision encoder, leaving the substantial latency of the visual encoding phase unoptimized. Second, under strict token budgets, these methods often fail to jointly preserve holistic visual contexts and fine-grained details, leading to performance degradation. To address these bottlenecks, we propose PACE (Pixel-Adaptive Condense and Extract), a training-free inference framework that accelerates both the vision encoder and the Large Language Model (LLM) via a unified Condense-and-Extract paradigm. During the Condense stage, an Adaptive Pixel Compressor (APC) evaluates visual information density prior to encoding, adaptively downsampling redundant inputs, curtailing encoder computation while preserving global context and essential visual cues. In the Extract stage, a Dynamic Dual-Attention Extractor (DDAE) selectively retains visual tokens via a fusion of internal visual signals from the encoder and semantic signals from the LLM, safeguarding task-critical details. By integrating PACE into Qwen2.5-VL-7B, the model retains 93.8% of its original performance while utilizing only 10% of the visual tokens, yielding a 3.1x speedup in time to first token (TTFT). Our code is available at https://github.com/jjL357/PACE.

cs.CV↗
arXiv2026

STRUCTUREDAGENT: Planning with AND/OR Trees for Long-Horizon Web Tasks

Existing LLM-based web agents struggle on complex, long-horizon tasks due to limited in-context memory, weak planning abilities, and greedy behaviors that lead to premature termination. To address these challenges, we propose \SA{}, a hierarchical planning framework that interleaves planning and execution via dynamic $\ANDOR$ trees. The framework separates structural planning from LLM-based reasoning, enabling principled error recovery through node repair, systematic exploration of alternatives via OR nodes, and modular plans that can facilitate human intervention. On WebArena (630 tasks), \SA{} achieves a $\sim$53\% success rate vs.\ $\sim$46\% for AgentOccam, and on complex multi-constraint Amazon shopping tasks, gains reach 10\% over the strongest baseline.

cs.AI↗
arXiv2026

Ripple-Pivot Search: Active Parallel Decoding for Diffusion Large Language Models

Diffusion Large Language Models (dLLMs) have emerged as a competitive alternative to autoregressive language models, offering the potential for substantially faster inference through parallel decoding. Existing parallel decoding schedulers typically commit positions only after they meet a per-position criterion, overlooking how early commitments may benefit subsequent decoding. We identify a ripple effect in dLLM decoding: proactively committing a mid-entropy pivot position can induce a pronounced reduction in uncertainty across the remaining masked positions. This uncertainty reduction allows subsequent steps to unmask more tokens in parallel, thereby accelerating the overall decoding process. To exploit the ripple effect, we propose Ripple-Pivot Search (RPS), a novel training-free decoding method that seeks mid-entropy positions as promising candidate pivots (where to decode), and determines their token assignment that yields the greatest downstream benefit via lookahead evaluation (what to decode). Across 3 dLLMs and 4 reasoning and code-generation benchmarks, RPS achieves 4-10$\times$ wall-clock speedup over the standard decoder while preserving generation quality, and improves accuracy over the previous lookahead baseline by up to 5.49% while delivering higher throughput in most settings. When integrated with KV caching, RPS further achieves up to 18$\times$ wall-clock speedup over the standard decoder.

cs.CL↗
arXiv2026

Commutator Geometry and Information Preservation in the Quantum Switch

Two commuting channels yield the same composite channel in either fixed order, yet their quantum switch can alter information preservation. We study how this effect depends on the input state, retaining the joint output of the control and target. Exact Kraus commutator identities determine overlap and squared fidelity gaps. Fixed order preserves product inputs at least as well as the switch in fidelity. The reverse inequality holds for pure states of two qubits whose control Schmidt basis can be chosen to coincide with the order basis. A commutator matrix determines the target basis that maximizes the squared fidelity gap at fixed entanglement within this aligned family. For Pauli $X$ and $Y$ channels with equal error probabilities, the switch replaces a logical phase flip with an operator that stabilizes a joint code. At fixed noise, the switch preserves at least as much distinguishability and quantum Fisher information as fixed order for every state family in this code. For the pure encoded family studied here, exact formulas connect the phase information gain to the squared fidelity gap and quantify the reduction in information about the noise probability. Optimizing the probes yields equal maximal quantum Fisher information for noise estimation under both joint channels. These results connect commutator geometry and logical error structure to state preservation and the retention of encoded information.

quant-ph↗
arXiv2026

Reasoning Quality Matters: Combating Reasoning Collapse in LLM-based Embedding Learning

Large Language Models (LLMs) have recently shown strong potential for producing context-rich text embeddings for retrieval. Most existing methods either treat embedding learning as passive feature extraction or exploit LLM reasoning through instruction following for better embedding optimization. However, specialization toward embedding objectives can suppress useful reasoning generation or produce retrieval-irrelevant text. We refer to these two forms of degradation as reasoning collapse. To address this issue, we propose CoFree (Collapse-Free Reasoning Embedding), a two-stage framework that progressively integrates LLM reasoning into query and document embedding optimization while preserving reasoning quality. At the first stage, CoFree applies reference-guided supervised fine-tuning to restore the reasoning ability and retain representational strength of the foundation embedding model. At the second stage, we introduce dual rewards, an embedding-oriented reward and a reasoning-oriented reward, to guarantee fine-grained reasoning of the relevance toward the embedding goal in reinforcement learning. This endpoint-coupled optimization transforms embedding learning from static alignment into a high-quality reasoning-guided search process for retrieval. Extensive experiments demonstrate the effectiveness of CoFree, with CoFree-4B achieving an average absolute improvement of 2.8 nDCG@10 points over Qwen3-Embedding-4B across 22 datasets from MTEB and BRIGHT. Online experiments in a real-world retrieval system further show consistent gains. Code, RTED, and model checkpoints will be made publicly available.

cs.IR↗
arXiv2026

Understanding Reliability in LLM-based Human Behavior Simulation

Large language models (LLMs) are increasingly used to simulate human survey responses and behavioral reactions, yet unreliable simulations can mislead social science conclusions. However, existing evaluations focus on end-to-end scores, leaving it unclear how different aspects of the simulation process interact to determine reliability. We propose ReliMap, which decomposes LLM-based human behavior simulation into three structured layers and evaluates reliability at both the individual level (R1) and population level (R2) across three configuration dimensions: model capacity, profile completeness, and population coverage. Through experiments across four simulation tasks and eleven LLMs, we find that all models exhibit substantial distributional bias without profile conditioning. Profile conditioning reduces this bias with diminishing returns. Larger models benefit more, and attribute informativeness matters more than quantity. Critically, R1 gains do not reliably transfer to R2--individual and population-level reliability can move in opposite directions. At the population layer, increasing coverage reduces variance but not systematic bias, with R2 stabilizing at around 50-100 individuals. These findings highlight that reliable simulation cannot be achieved by optimizing any single layer in isolation, but requires coordinated improvement across all three.

cs.CL↗
arXiv2026

ForgeTrain: Forging Production-Grade Training Frameworks via Harness-Driven AI Development

Training large models still relies on general-purpose frameworks such as Megatron-LM, whose generality tax constrains scenario-specific optimization and adds runtime overhead through accumulated abstraction. AI code generation reduces the cost of building a framework, and makes it affordable to forge one per scenario. We propose Forge Engineering: building a dedicated implementation from scratch for each scenario and iteratively optimizing it toward peak performance under correctness and usability constraints. Dedicated implementations inherit no abstraction boundaries, so they can integrate optimizations across the stack and reach a higher performance ceiling. We instantiate this paradigm for training frameworks as ForgeTrain, which holds a trusted framework as a golden reference and relaxes equivalence monotonically from Bit-for-Bit to Surpass. Experiments across multiple model--hardware configurations show that ForgeTrain consistently produces correct training engines and improves MFU over established training frameworks by 4.7--33.2%. To our knowledge this is the first production-grade training framework forged end-to-end by AI to match or surpass its human reference.

cs.SE↗
arXiv2026

ReUnit: Multi-Granularity Visual Unitization for Long Video Understanding

Long-video understanding is constrained by the limited visual input capacity of video multimodal large language models (Video-MLLMs). Existing methods mainly optimize which content to retain, while the presentation of retained content often remains fixed. As a result, the same balance between spatial detail and content coverage is imposed across the entire visual input. We propose ReUnit, a training-free and query-aware framework that jointly determines which content to retain and how it should be presented. Guided by frame-level query relevance, ReUnit constructs and allocates visual units at multiple presentation granularities. More relevant content receives finer presentation, while broader context is represented more compactly. It realizes these granularities using visual units that carry one, four, or nine source frames and are each rendered as a standard image. Across four benchmarks and visual input budgets from 4 to 32 units, ReUnit achieves the highest average score at every tested budget. It improves over Uniform Sampling by 8.3--10.1 points on average, and the gains transfer across three additional Video-MLLM families. Project resources are available at https://github.com/charon525/ReUnit.

cs.CV↗
arXiv2026

Preserving Subject-Clarity in Image Outpainting with Multiscale Wavelet Supervision

Commercial and advertising images are frequently affected by poor framing, partially cropped subjects, truncated text or logos, and insufficient context, all of which can reduce subject clarity, i.e., the ability of an image to clearly communicate its primary subject. Image outpainting offers a scalable solution by extending image boundaries and recovering missing content and context. However, existing diffusion-based outpainting methods often produce visually plausible completions while degrading subject fidelity through structural inconsistencies, semantic drift, or loss of fine-grained detail. To address this limitation, we propose a subject clarity outpainting framework that combines vision-language model (VLM)-guided semantic conditioning with multiscale wavelet supervision for subject-localized detail preservation. To support training, we develop a subject-centric data curation pipeline that constructs subject-intersecting outpainting pairs from advertising and natural images. The resulting objective introduces no additional inference cost and is designed to be compatible with diffusion-based backbones. Across four advertising and natural-image benchmarks, our method improves subject clarity, reducing subject-centered DreamSim error and FID on average by 3.0% and 2.4% over matched supervised fine-tuning, and by 10.8% and 7.7% over the strongest state-of-the-art approach per dataset, respectively.

cs.CV↗
arXiv2026

Quantum Change Interval: Exact Asymptotics for Minimum Error Localization

We study a returning quantum change interval in which a source emits $\lvertψ\rangle$ over one interval and $\lvert0\rangle$ elsewhere. A collective measurement on the full sequence identifies both endpoints with minimum error. We analyze the Gram matrix using Toeplitz comparison and Følner transfer, together with an exact decomposition by excitation number and interval hull. The resulting bounds establish asymptotic Bayes optimality of the square root measurement (SRM). Let $c=\lvert\langle0\vertψ\rangle\rvert$ and $p_1(x)=4(1-x^2)K^2(x^2)/π^2$, where $K$ is the complete elliptic integral of the first kind. For a known interval length $i$, the SRM success probability and the Bayes optimum converge to the same Toeplitz symbol integral as the number $N$ of translations grows. For fixed $i$ and $0<c<1$, their gap is $P_{\mathrm{opt}}(G_{N,i})-P_{\mathrm{SRM}}(G_{N,i})=O_{i,c}(N^{-1/2})$. If $i$ and $N$ both diverge, their common limit is $p_1(c^2)$, with no constraint on their relative growth. For unknown length, the uniform prior over all $M_n=n(n+1)/2$ nonempty intervals gives the common limit $p_1(c)^2$ at fixed overlap. For a varying overlap $c_n$, set $τ_n=n(1-c_n)(\log n)^2$. Uniformly for $0\leqτ_n\leq T$, we obtain $M_nP_X=(1+2\sqrt{τ_n}/π+\sqrt{2}τ_n/π^2)^2+O_T(\log\log n/\log n)$, where $X\in\{\mathrm{tr},\mathrm{SRM},\mathrm{opt}\}$. More generally, if $c_n$ approaches one from below and $n p_1(c_n)\to\infty$, the same three quantities satisfy $P_X\sim p_1(c_n)^2$. These asymptotic laws also extend to joint detection and exact localization in the presence of a no change prior.

quant-ph↗
arXiv2026

Pailitao-MMSearch: Building Native E-Commerce Multimodal Search Foundation

The evolution of e-commerce has fundamentally transformed how users search for products, shifting from simple text-based keyword queries to complex multimodal interactions that seamlessly combine product images, natural language descriptions, and mixed-intent instructions. However, existing approaches face a critical dilemma: single-modal specialist models, deployed independently for text retrieval, visual search, and voice recognition, operate in isolation and cannot handle cross-modal queries, while general-purpose vision-language models lack the domain-specific knowledge necessary for fine-grained product understanding, user behavior modeling, and commercial intent reasoning. In this work, we present Pailitao-MMSearch, one native e-commerce multimodal search foundation model designed to bridge this gap. Our approach introduces three key innovations: (1)HybSID (Hybrid Semantic ID);(2)a two-stage continual pre-training strategy; and (3)a hybrid reasoning post-training pipeline. Built upon Qwen and deployed on Taobao's Pailitao multimodal search platform, Pailitao-MMSearch achieves substantial improvements in online A/B testing, including up to +13.61\% in Gross Merchandise Volume (GMV) and +8.21\% in transaction volume compared to traditional multi-modal search pipeline, demonstrating the effectiveness of our native e-commerce multimodal search large language models.

cs.AI↗
arXiv2026

Grasp in Gaussians: Fast Monocular Reconstruction of Dynamic Hand-Object Interactions

We present Grasp in Gaussians (GraG), a fast and robust method for reconstructing dynamic 3D hand-object interactions from a single monocular video. Unlike recent approaches that optimize heavy neural representations, our method focuses on tracking the hand and the object efficiently, once initialized from pretrained large models. Our key insight is that, given strong pretrained object and hand priors, accurate and temporally stable hand-object motion can be recovered using a compact Sum-of-Gaussians (SoG) representation, revived from classical tracking literature and integrated with generative Gaussian-based initializations. We initialize object pose and geometry using a video-adapted SAM3D pipeline, then convert the resulting dense Gaussian representation into a lightweight SoG via subsampling. This compact representation enables efficient and fast tracking while preserving geometric fidelity, with the pretrained priors and simple geometric/contact losses providing accurate hand-object placement. For the hand, we adopt a complementary strategy: starting from off-the-shelf monocular hand pose initialization, we refine hand motion using simple yet effective 2D joint, silhouette, depth, and contact losses, avoiding per-frame refinement of a detailed 3D hand appearance model while maintaining stable articulation. Extensive experiments on public benchmarks demonstrate that GraG reconstructs temporally coherent hand-object interactions on long sequences 4.4x-38.9x faster than prior work while preserving temporally coherent hand-object motion.

cs.CV↗
arXiv2026

Spatio-Temporal Parallelism for Diffusion Model Inference on Heterogeneous Multi-GPU Systems

The widespread adoption of diffusion models for image generation necessitates efficient parallel inference to manage their substantial computational overhead. However, current parallel inference paradigms primarily target homogeneous clusters, often failing to maintain high efficiency in realistic, heterogeneous multi-GPU environments where hardware disparities and fluctuating background workloads cause severe straggler effects. This paper introduces Orchestra, a robust framework that orchestrates fine-grained parallelism across both temporal and spatial dimensions to harmonize computational loads in such settings. Temporally, Orchestra employs a novel computation-aware step allocator using a tiered step reduction strategy, intelligently pruning denoising steps on slower devices after warmup phases and execution synchronization. Spatially, Orchestra performs an elastic patch parallelism mechanism which adaptively adjusts the spatial workload intensity by assigning non-uniform image patches tailored to GPUs according to their computational capability. Extensive experiments on load-imbalanced and heterogeneous clusters validate Orchestra's efficacy in mitigating performance bottlenecks. Compared to patch parallelism, a state-of-the-art diffusion inference framework, our method reduces end-to-end latency by up to 45% and significantly boosts resource utilization on heterogeneous GPUs.

cs.DC↗