arXiv subjects

Xu Chen

Publications and source records attributed to Xu Chen.

At least 127 records · Page 7Linked to original sources

arXiv2026

Venus: An Efficient Edge Memory-and-Retrieval System for VLM-based Online Video Understanding

Vision-language models (VLMs) have demonstrated impressive multimodal comprehension capabilities and are being deployed in an increasing number of online video understanding applications. While recent efforts extensively explore advancing VLMs' reasoning power in these cases, deployment constraints are overlooked, leading to overwhelming system overhead in real-world deployments. To address that, we propose Venus, an on-device memory-and-retrieval system for efficient online video understanding. Venus proposes an edge-cloud disaggregated architecture that sinks memory construction and keyframe retrieval from cloud to edge, operating in two stages. In the ingestion stage, Venus continuously processes streaming edge videos via scene segmentation and clustering, where the selected keyframes are embedded with a multimodal embedding model to build a hierarchical memory for efficient storage and retrieval. In the querying stage, Venus indexes incoming queries from memory, and employs a threshold-based progressive sampling algorithm for keyframe selection that enhances diversity and adaptively balances system cost and reasoning accuracy. Our extensive evaluation shows that Venus achieves a 15x-131x speedup in total response latency compared to state-of-the-art methods, enabling real-time responses within seconds while maintaining comparable or even superior reasoning accuracy.

cs.DC↗

arXiv2026

Parallel Latent Reasoning for Sequential Recommendation

Capturing complex user preferences from sparse behavioral sequences remains a fundamental challenge in sequential recommendation. Recent latent reasoning methods have shown promise by extending test-time computation through multi-step reasoning, yet they exclusively rely on depth-level scaling along a single trajectory, suffering from diminishing returns as reasoning depth increases. To address this limitation, we propose \textbf{Parallel Latent Reasoning (PLR)}, a novel framework that pioneers width-level computational scaling by exploring multiple diverse reasoning trajectories simultaneously. PLR constructs parallel reasoning streams through learnable trigger tokens in continuous latent space, preserves diversity across streams via global reasoning regularization, and adaptively synthesizes multi-stream outputs through mixture-of-reasoning-streams aggregation. Extensive experiments on three real-world datasets demonstrate that PLR substantially outperforms state-of-the-art baselines while maintaining real-time inference efficiency. Theoretical analysis further validates the effectiveness of parallel reasoning in improving generalization capability. Our work opens new avenues for enhancing reasoning capacity in sequential recommendation beyond existing depth scaling.

cs.IR↗

arXiv2025

CoDrone: Autonomous Drone Navigation Assisted by Edge and Cloud Foundation Models

Autonomous navigation for Unmanned Aerial Vehicles faces key challenges from limited onboard computational resources, which restrict deployed deep neural networks to shallow architectures incapable of handling complex environments. Offloading tasks to remote edge servers introduces high latency, creating an inherent trade-off in system design. To address these limitations, we propose CoDrone - the first cloud-edge-end collaborative computing framework integrating foundation models into autonomous UAV cruising scenarios - effectively leveraging foundation models to enhance performance of resource-constrained unmanned aerial vehicle platforms. To reduce onboard computation and data transmission overhead, CoDrone employs grayscale imagery for the navigation model. When enhanced environmental perception is required, CoDrone leverages the edge-assisted foundation model Depth Anything V2 for depth estimation and introduces a novel one-dimensional occupancy grid-based navigation method - enabling fine-grained scene understanding while advancing efficiency and representational simplicity of autonomous navigation. A key component of CoDrone is a Deep Reinforcement Learning-based neural scheduler that seamlessly integrates depth estimation with autonomous navigation decisions, enabling real-time adaptation to dynamic environments. Furthermore, the framework introduces a UAV-specific vision language interaction module incorporating domain-tailored low-level flight primitives to enable effective interaction between the cloud foundation model and the UAV. The introduction of VLM enhances open-set reasoning capabilities in complex unseen scenarios. Experimental results show CoDrone outperforms baseline methods under varying flight speeds and network conditions, achieving a 40% increase in average flight distance and a 5% improvement in average Quality of Navigation.

cs.RO↗

arXiv2025

RecGPT-V2 Technical Report

Large language models (LLMs) have demonstrated remarkable potential in transforming recommender systems from implicit behavioral pattern matching to explicit intent reasoning. While RecGPT-V1 successfully pioneered this paradigm by integrating LLM-based reasoning into user interest mining and item tag prediction, it suffers from four fundamental limitations: (1) computational inefficiency and cognitive redundancy across multiple reasoning routes; (2) insufficient explanation diversity in fixed-template generation; (3) limited generalization under supervised learning paradigms; and (4) simplistic outcome-focused evaluation that fails to match human standards. To address these challenges, we present RecGPT-V2 with four key innovations. First, a Hierarchical Multi-Agent System restructures intent reasoning through coordinated collaboration, eliminating cognitive duplication while enabling diverse intent coverage. Combined with Hybrid Representation Inference that compresses user-behavior contexts, our framework reduces GPU consumption by 60% and improves exclusive recall from 9.39% to 10.99%. Second, a Meta-Prompting framework dynamically generates contextually adaptive prompts, improving explanation diversity by +7.3%. Third, constrained reinforcement learning mitigates multi-reward conflicts, achieving +24.1% improvement in tag prediction and +13.0% in explanation acceptance. Fourth, an Agent-as-a-Judge framework decomposes assessment into multi-step reasoning, improving human preference alignment. Online A/B tests on Taobao demonstrate significant improvements: +2.98% CTR, +3.71% IPV, +2.19% TV, and +11.46% NER. RecGPT-V2 establishes both the technical feasibility and commercial viability of deploying LLM-powered intent reasoning at scale, bridging the gap between cognitive exploration and industrial utility.

cs.IR↗

arXiv2025

Soul: Breathe Life into Digital Human for High-fidelity Long-term Multimodal Animation

We propose a multimodal-driven framework for high-fidelity long-term digital human animation termed $\textbf{Soul}$, which generates semantically coherent videos from a single-frame portrait image, text prompts, and audio, achieving precise lip synchronization, vivid facial expressions, and robust identity preservation. We construct Soul-1M, containing 1 million finely annotated samples with a precise automated annotation pipeline (covering portrait, upper-body, full-body, and multi-person scenes) to mitigate data scarcity, and we carefully curate Soul-Bench for comprehensive and fair evaluation of audio-/text-guided animation methods. The model is built on the Wan2.2-5B backbone, integrating audio-injection layers and multiple training strategies together with threshold-aware codebook replacement to ensure long-term generation consistency. Meanwhile, step/CFG distillation and a lightweight VAE are used to optimize inference efficiency, achieving an 11.4$\times$ speedup with negligible quality loss. Extensive experiments show that Soul significantly outperforms current leading open-source and commercial models on video quality, video-text alignment, identity preservation, and lip-synchronization accuracy, demonstrating broad applicability in real-world scenarios such as virtual anchors and film production. Project page at https://zhangzjn.github.io/projects/Soul/

cs.CV↗

arXiv2025

Enhancing Sequential Recommendation with World Knowledge from Large Language Models

Sequential Recommendation System~(SRS) has become pivotal in modern society, which predicts subsequent actions based on the user's historical behavior. However, traditional collaborative filtering-based sequential recommendation models often lead to suboptimal performance due to the limited information of their collaborative signals. With the rapid development of LLMs, an increasing number of works have incorporated LLMs' world knowledge into sequential recommendation. Although they achieve considerable gains, these approaches typically assume the correctness of LLM-generated results and remain susceptible to noise induced by LLM hallucinations. To overcome these limitations, we propose GRASP (Generation Augmented Retrieval with Holistic Attention for Sequential Prediction), a flexible framework that integrates generation augmented retrieval for descriptive synthesis and similarity retrieval, and holistic attention enhancement which employs multi-level attention to effectively employ LLM's world knowledge even with hallucinations and better capture users' dynamic interests. The retrieved similar users/items serve as auxiliary contextual information for the later holistic attention enhancement module, effectively mitigating the noisy guidance of supervision-based methods. Comprehensive evaluations on two public benchmarks and one industrial dataset reveal that GRASP consistently achieves state-of-the-art performance when integrated with diverse backbones. The code is available at: https://anonymous.4open.science/r/GRASP-SRS.

cs.IR↗

arXiv2025

Explore More, Learn Better: Parallel MLLM Embeddings under Mutual Information Minimization

Embedding models are a cornerstone of modern AI. Driven by Multimodal Large Language Models (MLLMs), they have made great progress in architecture and data curation, while the holistic paradigm is still limited to SSC, i.e., single input, singular embedding, contrastive supervision, which collapses rich, multifaceted inputs into monolithic embeddings and fails to fully exploit MLLM capabilities. In this paper, we tailor one Parallel Decoupling Framework (PDF) for multimodal embedding learning, by utilizing the proprietary steerability of MLLMs, i.e., their ability to flexibly generate quite differentiated response under explicit instructions. Concretely, PDF conditions a shared MLLM backbone on distinct, learnable prefixes to roll out multiple parallel paths for one input, then relies on these paths to obtain parallel embeddings. To promote full parallel diversity, we employ Mutual Information Minimization (MIM) as an explicit constraint, coupled with per-path contrastive supervision to maintain semantic alignment. Such dual-objectives force PDF to yield robust semantic coverage and a generalizable embedding space. Ultimately, the remarkable embedding space are accessible at inference via one single forward pass, incurring negligible computational overhead. We instantiate PDF on multiple MLLM backbones and prove its effectiveness on MMEB benchmark. Significant gains are consistently achieved across various resolutions and model sizes, e.g., boosting the VLM2Vec-LLaVA-1.6-LR model by a remarkable +8.9% (7B), while the VLM2Vec-Qwen2VL models by +4.2% (2B) and +3.1% (7B). In terms of efficiency, our 2B model surpasses its baseline by +2.6% using only half the computational budget.

cs.LG↗

arXiv2025

LLM-based Agents Suffer from Hallucinations: A Survey of Taxonomy, Methods, and Directions

Driven by the rapid advancements of Large Language Models (LLMs), LLM-based agents have emerged as powerful intelligent systems capable of human-like cognition, reasoning, and interaction. These agents are increasingly being deployed across diverse real-world applications, including student education, scientific research, and financial analysis. However, despite their remarkable potential, LLM-based agents remain vulnerable to hallucination issues, which can result in erroneous task execution and undermine the reliability of the overall system design. Addressing this critical challenge requires a deep understanding and a systematic consolidation of recent advances on LLM-based agents. To this end, we present the first comprehensive survey of hallucinations in LLM-based agents. By carefully analyzing the complete workflow of agents, we propose a new taxonomy that identifies different types of agent hallucinations occurring at different stages. Furthermore, we conduct an in-depth examination of eighteen triggering causes underlying the emergence of agent hallucinations. Through a detailed review of a large number of existing studies, we summarize approaches for hallucination mitigation and detection, and highlight promising directions for future research. We hope this survey will inspire further efforts toward addressing hallucinations in LLM-based agents, ultimately contributing to the development of more robust and reliable agent systems.

cs.AI↗

arXiv2025

Leveraging LLM-based agents for social science research: insights from citation network simulations

The emergence of Large Language Models (LLMs) demonstrates their potential to encapsulate the logic and patterns inherent in human behavior simulation by leveraging extensive web data pre-training. However, the boundaries of LLM capabilities in social simulation remain unclear. To further explore the social attributes of LLMs, we introduce the CiteAgent framework, designed to generate citation networks based on human-behavior simulation with LLM-based agents. CiteAgent successfully captures predominant phenomena in real-world citation networks, including power-law distribution, citational distortion, and shrinking diameter. Building on this realistic simulation, we establish two LLM-based research paradigms in social science: LLM-SE (LLM-based Survey Experiment) and LLM-LE (LLM-based Laboratory Experiment). These paradigms facilitate rigorous analyses of citation network phenomena, allowing us to validate and challenge existing theories. Additionally, we extend the research scope of traditional science of science studies through idealized social experiments, with the simulation experiment results providing valuable insights for real-world academic environments. Our work demonstrates the potential of LLMs for advancing science of science research in social science.

physics.soc-ph↗

arXiv2025

OpenUS: A Fully Open-Source Foundation Model for Ultrasound Image Analysis via Self-Adaptive Masked Contrastive Learning

Ultrasound (US) is one of the most widely used medical imaging modalities, thanks to its low cost, portability, real-time feedback, and absence of ionizing radiation. However, US image interpretation remains highly operator-dependent and varies significantly across anatomical regions, acquisition protocols, and device types. These variations, along with unique challenges such as speckle, low contrast, and limited standardized annotations, hinder the development of generalizable, label-efficient ultrasound AI models. In this paper, we propose OpenUS, the first reproducible, open-source ultrasound foundation model built on a large collection of public data. OpenUS employs a vision Mamba backbone, capturing both local and global long-range dependencies across the image. To extract rich features during pre-training, we introduce a novel self-adaptive masking framework that combines contrastive learning with masked image modeling. This strategy integrates the teacher's attention map with student reconstruction loss, adaptively refining clinically-relevant masking to enhance pre-training effectiveness. OpenUS also applies a dynamic learning schedule to progressively adjust the difficulty of the pre-training process. To develop the foundation model, we compile the largest to-date public ultrasound dataset comprising over 308K images from 42 publicly available datasets, covering diverse anatomical regions, institutions, imaging devices, and disease types. Our pre-trained OpenUS model can be easily adapted to specific downstream tasks by serving as a backbone for label-efficient fine-tuning. Code is available at https://github.com/XZheng0427/OpenUS.

cs.CV↗

arXiv2025

CoEdge-RAG: Optimizing Hierarchical Scheduling for Retrieval-Augmented LLMs in Collaborative Edge Computing

Motivated by the imperative for real-time responsiveness and data privacy preservation, large language models (LLMs) are increasingly deployed on resource-constrained edge devices to enable localized inference. To improve output quality, retrieval-augmented generation (RAG) is an efficient technique that seamlessly integrates local data into LLMs. However, existing edge computing paradigms primarily focus on single-node optimization, neglecting opportunities to holistically exploit distributed data and heterogeneous resources through cross-node collaboration. To bridge this gap, we propose CoEdge-RAG, a hierarchical scheduling framework for retrieval-augmented LLMs in collaborative edge computing. In general, privacy constraints preclude accurate a priori acquisition of heterogeneous data distributions across edge nodes, directly impeding RAG performance optimization. Thus, we first design an online query identification mechanism using proximal policy optimization (PPO), which autonomously infers query semantics and establishes cross-domain knowledge associations in an online manner. Second, we devise a dynamic inter-node scheduling strategy that balances workloads across heterogeneous edge nodes by synergizing historical performance analytics with real-time resource thresholds. Third, we develop an intra-node scheduler based on online convex optimization, adaptively allocating query processing ratios and memory resources to optimize the latency-quality trade-off under fluctuating assigned loads. Comprehensive evaluations across diverse QA benchmarks demonstrate that our proposed method significantly boosts the performance of collaborative retrieval-augmented LLMs, achieving performance gains of 4.23\% to 91.39\% over baseline methods across all tasks.

cs.DC↗

arXiv2025

Wave-Particle (Continuous-Discrete) Dualistic Visual Tokenization for Unified Understanding and Generation

The unification of understanding and generation within a single multi-modal large model (MLLM) remains one significant challenge, largely due to the dichotomy between continuous and discrete visual tokenizations. Continuous tokenizer (CT) achieves strong performance by bridging multiple independently-trained understanding modules and generation modules, but suffers from complex multi-stage pipelines and substantial engineering overhead. Conversely, discrete tokenizers (DT) offer a conceptually elegant idea by quantizing each image into a primitive, but inevitably leading to information loss and performance degradation. To resolve this tension, we question the binary choice between CT and DT, inspired by the wave-particle duality of light, and propose the Continuous-Discrete Dualistic Visual Tokenizer (CDD-VT). We treat visual data as a flexible composition of image primitives derived from quantized codebooks, with the crucial insight that the primitive number assigned to each visual sample is adaptively determined according to its complexity: simple instances use a few primitives, emulating discrete tokenization, while complex instances use many, approximating continuous tokenization. Two core components are designed: Diverse Quantitative Primitives, which encourage primitives orthogonality to better populate information space, and Dynamic Primitive Allocator, which assesses sample complexity to determine the optimal set of primitives. Extensive experiments on reconstruction, retrieval and classification show that CDD-VT achieves superior performance over to specialized CT and DT, effectively getting strong result within a concise and scalable MLLM.

cs.CV↗

arXiv2025

SAIL-Embedding Technical Report: Omni-modal Embedding Foundation Model

Multimodal embedding models aim to yield informative unified representations that empower diverse cross-modal tasks. Despite promising developments in the evolution from CLIP-based dual-tower architectures to large vision-language models, prior works still face unavoidable challenges in real-world applications and business scenarios, such as the limited modality support, unstable training mechanisms, and industrial domain gaps. In this work, we introduce SAIL-Embedding, an omni-modal embedding foundation model that addresses these issues through tailored training strategies and architectural design. In the optimization procedure, we propose a multi-stage training scheme to boost the multifaceted effectiveness of representation learning. Specifically, the content-aware progressive training aims to enhance the model's adaptability to diverse downstream tasks and master enriched cross-modal proficiency. The collaboration-aware recommendation enhancement training further adapts multimodal representations for recommendation scenarios by distilling knowledge from sequence-to-item and ID-to-item embeddings while mining user historical interests. Concurrently, we develop the stochastic specialization and dataset-driven pattern matching to strengthen model training flexibility and generalizability. Experimental results show that SAIL-Embedding achieves SOTA performance compared to other methods in different retrieval tasks. In online experiments across various real-world scenarios integrated with our model, we observe a significant increase in Lifetime (LT), which is a crucial indicator for the recommendation experience. For instance, the model delivers the 7-day LT gain of +0.5% in the Douyin-Selected scenario. For the Douyin feed rank model, the match features produced by SAIL-Embedding yield a +0.1% AUC gain.

cs.IR↗

arXiv2025

Beyond Benchmarks: The Economics of AI Inference

The inference cost of Large Language Models (LLMs) has become a critical factor in determining their commercial viability and widespread adoption. This paper introduces a quantitative ``economics of inference'' framework, treating the LLM inference process as a compute-driven intelligent production activity. We analyze its marginal cost, economies of scale, and quality of output under various performance configurations. Based on empirical data from WiNEval-3.0, we construct the first ``LLM Inference Production Frontier,'' revealing three principles: diminishing marginal cost, diminishing returns to scale, and an optimal cost-effectiveness zone. This paper not only provides an economic basis for model deployment decisions but also lays an empirical foundation for the future market-based pricing and optimization of AI inference resources.

cs.AI↗

arXiv2025

Squeezed Light Generation in Periodically Poled Thin-Film Lithium Niobate Waveguides

Squeezed states of light play a key role in quantum-enhanced sensing and continuous-variable quantum information processing. Realizing integrated squeezed light sources is crucial for developing compact and scalable photonic quantum systems. In this work, we demonstrate on-chip broadband vacuum squeezing at telecommunication wavelengths on the thin-film lithium niobate (TFLN) platform. Our device integrates periodically poled lithium niobate (PPLN) nanophotonic waveguides with low-loss edge couplers, comprising bilayer inverse tapers and an SU-8 polymer waveguide. This configuration achieves a fiber-to-chip coupling loss of 1.4 dB and a total homodyne detection loss of 4 dB, enabling a measured squeezing level of 1.4 dB. Additional measurements in a more efficient PPLN waveguide (without low-loss couplers) infer an on-chip squeezing level of over 10 dB at a pump power of 62 mW. These results underscore the potential of TFLN platform for efficient and scalable squeezed light generation.

physics.optics↗

arXiv2025

Energy-Efficient and Dequantization-Free Q-LLMs: A Spiking Neural Network Approach to Salient Value Mitigation

In the era of large language models (LLMs), weight-activation quantization helps fit models on edge device by reducing memory and compute bit-widths. However, three challenges persist for energy constrained hardware: (1) even after quantization, multiply-accumulate (MAC) operations remain unavoidable and continue to dominate energy consumption; (2) dequantization (or per-tensor/channel rescaling) introduces extra arithmetic and data movement, increasing latency and energy; (3) uniform parameters bit widths clip salient values-while intra-channel mixed precision is generally impractical on current matrix hardware and memory. In contrast, brain-inspired Spiking Neural Networks (SNNs), owing to their binary spike-based information representation and the Integrate-and-Fire (IF) paradigm, naturally support mixed-precision storage and energy-efficient computation by replacing complex MACs with temporal Accumulate (ACCs). Motivated by this property, we propose SpikeQuant, which selectively applies mixed-precision quantization to activations with salient values and re-encodes them into binary spike counts, thereby enabling dynamic mixed storage of different bitwidths. Furthermore, by embedding the quantization scale into the threshold of the IF mechanism, our approach performs energy-efficient linear transformations on weights and activations while avoiding explicit dequantization. Experimental results demonstrate that SpikeQuant consistently achieves near-FP16 perplexity under W4A4 quantization while reducing energy cost by up to 4.6 times compared to existing methods, highlighting its effectiveness for accurate and energy-efficient LLM deployment.

cs.LG↗

arXiv2025

Dual Data Alignment Makes AI-Generated Image Detector Easier Generalizable

Existing detectors are often trained on biased datasets, leading to the possibility of overfitting on non-causal image attributes that are spuriously correlated with real/synthetic labels. While these biased features enhance performance on the training data, they result in substantial performance degradation when applied to unbiased datasets. One common solution is to perform dataset alignment through generative reconstruction, matching the semantic content between real and synthetic images. However, we revisit this approach and show that pixel-level alignment alone is insufficient. The reconstructed images still suffer from frequency-level misalignment, which can perpetuate spurious correlations. To illustrate, we observe that reconstruction models tend to restore the high-frequency details lost in real images (possibly due to JPEG compression), inadvertently creating a frequency-level misalignment, where synthetic images appear to have richer high-frequency content than real ones. This misalignment leads to models associating high-frequency features with synthetic labels, further reinforcing biased cues. To resolve this, we propose Dual Data Alignment (DDA), which aligns both the pixel and frequency domains. Moreover, we introduce two new test sets: DDA-COCO, containing DDA-aligned synthetic images for testing detector performance on the most aligned dataset, and EvalGEN, featuring the latest generative models for assessing detectors under new generative architectures such as visual auto-regressive generators. Finally, our extensive evaluations demonstrate that a detector trained exclusively on DDA-aligned MSCOCO could improve across 8 diverse benchmarks by a non-trivial margin, showing a +7.2% on in-the-wild benchmarks, highlighting the improved generalizability of unbiased detectors. Our code is available at: https://github.com/roy-ch/Dual-Data-Alignment.

cs.CV↗

arXiv2025

Cog-Rethinker: Hierarchical Metacognitive Reinforcement Learning for LLM Reasoning

Contemporary progress in large language models (LLMs) has revealed notable inferential capacities via reinforcement learning (RL) employing verifiable reward, facilitating the development of O1 and R1-like reasoning models. Directly training from base models with RL is called zero-RL. However, previous works rely upon activating LLMs' inherent capacities through fixed prompt templates. This strategy introduces substantial sampling inefficiencies for weak LLMs, as the majority of problems generate invalid outputs during accuracy-driven filtration in reasoning tasks, which causes a waste of samples. To solve this issue, we propose Cog-Rethinker, a novel hierarchical metacognitive RL framework for LLM reasoning. Our Cog-Rethinker mainly focuses on the rollout procedure in RL training. After the direct rollout, our Cog-Rethinker improves sample utilization in a hierarchical metacognitive two-stage framework. By leveraging human cognition during solving problems, firstly, it prompts policy to decompose zero-accuracy problems into subproblems to produce final reasoning results. Secondly, with zero-accuracy problems in previous rollout stage, it further prompts policy to refine these answers by referencing previous wrong solutions. Moreover, to enable cold-start of the two new reasoning patterns and maintain train-test consistency across prompt templates, our Cog-Rethinker applies supervised fine-tuning on the policy using correct samples of the two stages with direct rollout template. Experimental results demonstrate Cog-Rethinker's superior performance on various mathematical reasoning benchmarks, we also analyzed its improved sample efficiency that accelerates convergence compared to baseline methods.

cs.LG↗