arXiv subjects

Li Zhang

Publications and source records attributed to Li Zhang.

At least 235 records · Page 13Linked to original sources

arXiv2026

Thinking Longer, Not Always Smarter: Evaluating LLM Capabilities in Hierarchical Legal Reasoning

Case-based reasoning is a cornerstone of U.S. legal practice, requiring professionals to argue about a current case by drawing analogies to and distinguishing from past precedents. While Large Language Models (LLMs) have shown remarkable capabilities, their proficiency in this complex, nuanced form of reasoning needs further investigation. We propose a formal framework that decomposes the process of identifying significant distinctions between cases into three-stage reasoning tasks. Our framework models cases using factual predicates called factors, organizes them into a legal knowledge hierarchy, and defines verifiable rules for identifying distinctions, analyzing their argumentative support, and evaluating their significance. Through comprehensive evaluation of modern reasoning LLMs, we reveal a paradox: while models achieve high accuracy on surface-level reasoning (Task 1), performance degrades on hierarchical reasoning (Task 2: 64.82%-92.09%) and collapses on integrated analysis (Task 3: 11.46%-33.99%). Most strikingly, we find that models consistently expend more computational resources on incorrect responses than correct ones, suggesting that "thinking longer" does not always mean "thinking smarter." Our work provides a methodology for fine-grained analysis of LLM reasoning capabilities in complex domains and reveals fundamental limitations that must be addressed for robust and trustworthy legal AI.

cs.CL↗

arXiv2026

From Flatland to Space: Teaching Vision-Language Models to Perceive and Reason in 3D

Recent advances in LVLMs have improved vision-language understanding, but they still struggle with spatial perception, limiting their ability to reason about complex 3D scenes. Unlike previous approaches that incorporate 3D representations into models to improve spatial understanding, we aim to unlock the potential of VLMs by leveraging spatially relevant image data. To this end, we introduce a novel 2D spatial data generation and annotation pipeline built upon scene data with 3D ground-truth. This pipeline enables the creation of a diverse set of spatial tasks, ranging from basic perception tasks to more complex reasoning tasks. Leveraging this pipeline, we construct SPAR-7M, a large-scale dataset generated from thousands of scenes across multiple public datasets. In addition, we introduce SPAR-Bench, a benchmark designed to offer a more comprehensive evaluation of spatial capabilities compared to existing spatial benchmarks, supporting both single-view and multi-view inputs. Training on both SPAR-7M and large-scale 2D datasets enables our models to achieve state-of-the-art performance on 2D spatial benchmarks. Further fine-tuning on 3D task-specific datasets yields competitive results, underscoring the effectiveness of our dataset in enhancing spatial reasoning.

cs.CV↗

arXiv2026

SGDrive: Scene-to-Goal Hierarchical World Cognition for Autonomous Driving

Recent end-to-end autonomous driving approaches have leveraged Vision-Language Models (VLMs) to enhance planning capabilities in complex driving scenarios. However, VLMs are inherently trained as generalist models, lacking specialized understanding of driving-specific reasoning in 3D space and time. When applied to autonomous driving, these models struggle to establish structured spatial-temporal representations that capture geometric relationships, scene context, and motion patterns critical for safe trajectory planning. To address these limitations, we propose SGDrive, a novel framework that explicitly structures the VLM's representation learning around driving-specific knowledge hierarchies. Built upon a pre-trained VLM backbone, SGDrive decomposes driving understanding into a scene-agent-goal hierarchy that mirrors human driving cognition: drivers first perceive the overall environment (scene context), then attend to safety-critical agents and their behaviors, and finally formulate short-term goals before executing actions. This hierarchical decomposition provides the structured spatial-temporal representation that generalist VLMs lack, integrating multi-level information into a compact yet comprehensive format for trajectory planning. Extensive experiments on the NAVSIM benchmark demonstrate that SGDrive achieves state-of-the-art performance among camera-only methods on both PDMS and EPDMS, validating the effectiveness of hierarchical knowledge structuring for adapting generalist VLMs to autonomous driving.

cs.CV↗

arXiv2026

Language as Prior, Vision as Calibration: Metric Scale Recovery for Monocular Depth Estimation

Relative-depth foundation models transfer well, yet monocular metric depth remains ill-posed due to unidentifiable global scale and heightened domain-shift sensitivity. Under a frozen-backbone calibration setting, we recover metric depth via an image-specific affine transform in inverse depth and train only lightweight calibration heads while keeping the relative-depth backbone and the CLIP text encoder fixed. Since captions provide coarse but noisy scale cues that vary with phrasing and missing objects, we use language to predict an uncertainty-aware envelope that bounds feasible calibration parameters in an unconstrained space, rather than committing to a text-only point estimate. We then use pooled multi-scale frozen visual features to select an image-specific calibration within this envelope. During training, a closed-form least-squares oracle in inverse depth provides per-image supervision for learning the envelope and the selected calibration. Experiments on NYUv2 and KITTI improve in-domain accuracy, while zero-shot transfer to SUN-RGBD and DDAD demonstrates improved robustness over strong language-only baselines.

cs.CV↗

arXiv2026

Sparsity-Aware Streaming SNN Accelerator with Output-Channel Dataflow for Automatic Modulation Classification

The rapid advancement of wireless communication technologies, including 5G, emerging 6G networks, and the large-scale deployment of the Internet of Things (IoT), has intensified the need for efficient spectrum utilization. Automatic modulation classification (AMC) plays a vital role in cognitive radio systems by enabling real-time identification of modulation schemes for dynamic spectrum access and interference mitigation. While deep neural networks (DNNs) offer high classification accuracy, their computational and energy demands pose challenges for real-time edge deployment. Spiking neural networks (SNNs), with their event-driven nature, offer inherent energy efficiency, but achieving both high throughput and low power under constrained hardware resources remains challenging. This work proposes a sparsity-aware SNN streaming accelerator optimized for AMC tasks. Unlike traditional systolic arrays that exploit sparsity but suffer from low throughput, or streaming architectures that achieve high throughput but cannot fully utilize input and weight sparsity, our design integrates both advantages. By leveraging the fixed nature of kernels during inference, we apply the gated one-to-all product (GOAP) algorithm to compute only on non-zero input-weight intersections. Extra or empty iterations are precomputed and embedded into the inference dataflow, eliminating dynamic data fetches and enabling fully pipelined, control-free inter-layer execution. Implemented on an FPGA, our sparsity-aware output-channel dataflow streaming (SAOCDS) accelerator achieves 23.5 MS/s (approximately double the baseline throughput) on the RadioML 2016 dataset, while reducing dynamic power and maintaining comparable classification accuracy. These results demonstrate strong potential for real-time, low-power deployment in edge cognitive radio systems.

cs.AR↗

arXiv2026

CodeMEM: AST-Guided Adaptive Memory for Repository-Level Iterative Code Generation

Large language models (LLMs) substantially enhance developer productivity in repository-level code generation through interactive collaboration. However, as interactions progress, repository context must be continuously preserved and updated to integrate newly validated information. Meanwhile, the expanding session history increases cognitive burden, often leading to forgetting and the reintroduction of previously resolved errors. Existing memory management approaches show promise but remain limited by natural language-centric representations. To overcome these limitations, we propose CodeMEM, an AST-guided dynamic memory management system tailored for repository-level iterative code generation. Specifically, CodeMEM introduces the Code Context Memory component that dynamically maintains and updates repository context through AST-guided LLM operations, along with the Code Session Memory that constructs a code-centric representation of interaction history and explicitly detects and mitigates forgetting through AST-based analysis. Experimental results on the instruction-following benchmark CodeIF-Bench and the code generation benchmark CoderEval demonstrate that CodeMEM achieves state-of-the-art performance, improving instruction following by 12.2% for the current turn and 11.5% for the session level, and reducing interaction rounds by 2-3, while maintaining competitive inference latency and token efficiency.

cs.SE↗

arXiv2026

Empowering Small Language Models with Factual Hallucination-Aware Reasoning for Financial Classification

Small language models (SLMs) are increasingly used for financial classification due to their fast inference and local deployability. However, compared with large language models, SLMs are more prone to factual hallucinations in reasoning and exhibit weaker classification performance. This raises a natural question: Can mitigating factual hallucinations improve SLMs' financial classification? To address this, we propose a three-step pipeline named AAAI (Association Identification, Automated Detection, and Adaptive Inference). Experiments on three representative SLMs reveal that: (1) factual hallucinations are positively correlated with misclassifications; (2) encoder-based verifiers effectively detect factual hallucinations; and (3) incorporating feedback on factual errors enables SLMs' adaptive inference that enhances classification performance. We hope this pipeline contributes to trustworthy and effective applications of SLMs in finance.

cs.AI↗

arXiv2025

5-GHz chip-based quantum key distribution with 1Mbps secure key rate over 150 km

Quantum key distribution (QKD) enables secure communication by harnessing the fundamental principles of quantum physics, which inherently guarantee information-theoretic security and intrinsic resistance to quantum computing attacks. However, the secure key rate of QKD typically decreases exponentially with increasing channel distance. In this work, by developing a novel polarization-state preparation method, an ultra-low time-jitter laser source and superconducting nanowire single-photon detectors, we demonstrate a 5-GHz integrated QKD system featuring ultra-low quantum bit error rates (QBERs). The system achieves secure key rates of 1.076 Mbps at 150 km and 105 kbps at 200 km over standard single-mode fiber channels, respectively. Our system substantially enhances the secure key rate, enabling high-resolution video calls with one-time-pad encryption over intercity backbone QKD links. This work represents a significant step forward in the development of high-performance practical QKD systems.

quant-ph↗

arXiv2025

Quantifying the Impact of Structured Output Format on Large Language Models through Causal Inference

Structured output from large language models (LLMs) has enhanced efficiency in processing generated information and is increasingly adopted in industrial applications. Prior studies have investigated the impact of structured output on LLMs' generation quality, often presenting one-way findings. Some suggest that structured format enhances completeness and factual accuracy, while others argue that it restricts the reasoning capacity of LLMs and leads to reductions in standard evaluation metrics. Potential limitations of these assessments include restricted testing scenarios, weakly controlled comparative settings, and reliance on coarse metrics. In this work, we present a refined analysis using causal inference. Based on one assumed and two guaranteed constraints, we derive five potential causal structures characterizing the influence of structured output on LLMs' generation: (1) collider without m-bias, (2) collider with m-bias, (3) single cause from instruction, (4) single cause from output format, and (5) independence. Across seven public and one developed reasoning tasks, we find that coarse metrics report positive, negative, or neutral effects of structured output on GPT-4o's generation. However, causal inference reveals no causal impact in 43 out of 48 scenarios. In the remaining 5, 3 involve multifaceted causal structures influenced by concrete instructions. Further experiments show that OpenAI-o3 are more resilient to output formats than general-purpose GPT-4o and GPT-4.1, highlighting an unaware advantage of reasoning models.

cs.CL↗

arXiv2025

A Tri-Dynamic Preprocessing Framework for UGC Video Compression

In recent years, user generated content (UGC) has become the dominant force in internet traffic. However, UGC videos exhibit a higher degree of variability and diverse characteristics compared to traditional encoding test videos. This variance challenges the effectiveness of data-driven machine learning algorithms for optimizing encoding in the broader context of UGC scenarios. To address this issue, we propose a Tri-Dynamic Preprocessing framework for UGC. Firstly, we employ an adaptive factor to regulate preprocessing intensity. Secondly, an adaptive quantization level is employed to fine-tune the codec simulator. Thirdly, we utilize an adaptive lambda tradeoff to adjust the rate-distortion loss function. Experimental results on large-scale test sets demonstrate that our method attains exceptional performance.

cs.MM↗

arXiv2025

Aligning Academia with Industry: An Empirical Study of Industrial Needs and Academic Capabilities in AI-Driven Software Engineering

The rapid advancement of large language models (LLMs) is fundamentally reshaping software engineering (SE), driving a paradigm shift in both academic research and industrial practice. While top-tier SE venues continue to show sustained or emerging focus on areas like automated testing and program repair, with researchers worldwide reporting continuous performance gains, the alignment of these academic advances with real industrial needs remains unclear. To bridge this gap, we first conduct a systematic analysis of 1,367 papers published in FSE, ASE, and ICSE between 2022 and 2025, identifying key research topics, commonly used benchmarks, industrial relevance, and open-source availability. We then carry out an empirical survey across 17 organizations, collecting 282 responses on six prominent topics, i.e., program analysis, automated testing, code generation/completion, issue resolution, pre-trained code models, and dependency management, through structured questionnaires. By contrasting academic capabilities with industrial feedback, we derive seven critical implications, highlighting under-addressed challenges in software requirements and architecture, the reliability and explainability of intelligent SE approaches, input assumptions in academic research, practical evaluation tensions, and ethical considerations. This study aims to refocus academic attention on these important yet under-explored problems and to guide future SE research toward greater industrial impact.

cs.SE↗

arXiv2025

Audio-Visual Cross-Modal Compression for Generative Face Video Coding

Generative face video coding (GFVC) is vital for modern applications like video conferencing, yet existing methods primarily focus on video motion while neglecting the significant bitrate contribution of audio. Despite the well-established correlation between audio and lip movements, this cross-modal coherence has not been systematically exploited for compression. To address this, we propose an Audio-Visual Cross-Modal Compression (AVCC) framework that jointly compresses audio and video streams. Our framework extracts motion information from video and tokenizes audio features, then aligns them through a unified audio-video diffusion process. This allows synchronized reconstruction of both modalities from a shared representation. In extremely low-rate scenarios, AVCC can even reconstruct one modality from the other. Experiments show that AVCC significantly outperforms the Versatile Video Coding (VVC) standard and state-of-the-art GFVC schemes in rate-distortion performance, paving the way for more efficient multimodal communication systems.

eess.IV↗

arXiv2025

Generative Preprocessing for Image Compression with Pre-trained Diffusion Models

Preprocessing is a well-established technique for optimizing compression, yet existing methods are predominantly Rate-Distortion (R-D) optimized and constrained by pixel-level fidelity. This work pioneers a shift towards Rate-Perception (R-P) optimization by, for the first time, adapting a large-scale pre-trained diffusion model for compression preprocessing. We propose a two-stage framework: first, we distill the multi-step Stable Diffusion 2.1 into a compact, one-step image-to-image model using Consistent Score Identity Distillation (CiD). Second, we perform a parameter-efficient fine-tuning of the distilled model's attention modules, guided by a Rate-Perception loss and a differentiable codec surrogate. Our method seamlessly integrates with standard codecs without any modification and leverages the model's powerful generative priors to enhance texture and mitigate artifacts. Experiments show substantial R-P gains, achieving up to a 30.13% BD-rate reduction in DISTS on the Kodak dataset and delivering superior subjective visual quality.

eess.IV↗

arXiv2025

A Preprocessing Framework for Video Machine Vision under Compression

There has been a growing trend in compressing and transmitting videos from terminals for machine vision tasks. Nevertheless, most video coding optimization method focus on minimizing distortion according to human perceptual metrics, overlooking the heightened demands posed by machine vision systems. In this paper, we propose a video preprocessing framework tailored for machine vision tasks to address this challenge. The proposed method incorporates a neural preprocessor which retaining crucial information for subsequent tasks, resulting in the boosting of rate-accuracy performance. We further introduce a differentiable virtual codec to provide constraints on rate and distortion during the training stage. We directly apply widely used standard codecs for testing. Therefore, our solution can be easily applied to real-world scenarios. We conducted extensive experiments evaluating our compression method on two typical downstream tasks with various backbone networks. The experimental results indicate that our approach can save over 15% of bitrate compared to using only the standard codec anchor version.

cs.MM↗

arXiv2025

Template-Guided Reconstruction of Pulmonary Segments with Neural Implicit Functions

High-quality 3D reconstruction of pulmonary segments plays a crucial role in segmentectomy and surgical planning for the treatment of lung cancer. Due to the resolution requirement of the target reconstruction, conventional deep learning-based methods often suffer from computational resource constraints or limited granularity. Conversely, implicit modeling is favored due to its computational efficiency and continuous representation at any resolution. We propose a neural implicit function-based method to learn a 3D surface to achieve anatomy-aware, precise pulmonary segment reconstruction, represented as a shape by deforming a learnable template. Additionally, we introduce two clinically relevant evaluation metrics to comprehensively assess the quality of the reconstruction. Furthermore, to address the lack of publicly available shape datasets for benchmarking reconstruction algorithms, we developed a shape dataset named Lung3D, which includes the 3D models of 800 labeled pulmonary segments and their corresponding airways, arteries, veins, and intersegmental veins. We demonstrate that the proposed approach outperforms existing methods, providing a new perspective for pulmonary segment reconstruction. Code and data will be available at https://github.com/HINTLab/ImPulSe.

cs.GR↗

arXiv2025

Patterns of X-ray and $γ$-ray Flares: from Blazar to Maganetar and Sun

Using Gaussian process methods, we analyzed the light curves of three extreme solar X-ray flares observed by the RHESSI satellite. Their variability characteristics were then compared with those of HXMT-HE X-ray burst (XRB; in SGR 1935+2154) associated with fast radio burst (FRB) 200428 and blazar $γ$-ray giant flares, to investigate the origins of these extreme flaring events. The variability patterns of the solar X-ray flares follow the stochastically driven damped simple harmonic oscillator (SHO) model. The derived timescales $t_{\rm B\underline{} steep}$ and $t_{\rm B\underline{~} flat}$ (corresponding to PSD breaks) are in the range of 4-7 s and 16-53 s, respectively. The FRB-associated HXMT-HE burst has a $Q$ value near 0.3, matching those of the solar flares occurred on 23 July 2002 (flare 1) and 3 November 2003 (flare 2). By contrast, blazar $γ$-ray giant flares show $Q >$ 0.3, similar to the solar flare occurred on 25 February 2014 (flare 3). We proposed that the critically damped state of the system may be the condition triggering the association between the XRB in SGR 1935+2154 and the FRB. In this scenario, the critical damping $Q$ value of the system is around 0.3, not the theoretical 0.5. The similarity in $Q$ values might imply that the FRB-associated HXMT-HE XRB and solar X-ray flares 1 $\&$ 2 share comparable dynamic behavior, while blazar $γ$-ray flares and solar X-ray flare 3 exhibit another distinct but similar dynamic behavior. Like solar X-ray flares, these extreme flares may all be related to the magnetic reconnection process.

astro-ph.HE↗

arXiv2025

A Mass-Independent Damping Timescale in Black Hole Accretion Systems

The scaling laws reveal the underlying structural similarities shared by astrophysical systems across vastly different scales. In black hole accretion systems, the scaling relations between the characteristic damping timescales (CDTs) of light curves and black hole mass offer valuable insights into the underlying physical structure of accretion disks. Here, we investigate the long-term hard X-ray CDTs of 106 black hole and neutron star accretion systems using light curves from the \textit{Swift} Burst Alert Telescope 157-month catalog. Unexpectedly, for the first time, we discover a mass-independent CDT in these black hole accretion systems, in contrast to well-established scaling laws. This puzzling phenomenon can be attributed to conductive timescales arising from disk--corona interactions, instead of the intrinsic accretion disk processes characterized by scaling laws, and it may further modulate jet emission in blazars. This result demonstrates thermal conduction as a key mechanism driving hard X-ray variability and offers new observational evidence for the disk--corona--jet connection.

astro-ph.HE↗

arXiv2025

Decoding Human-LLM Collaboration in Coding: An Empirical Study of Multi-Turn Conversations in the Wild

Large language models (LLMs) are increasingly acting as dynamic conversational interfaces, supporting multi-turn interactions that mimic human-like conversation and facilitate complex tasks like coding. While datasets such as LMSYS-Chat-1M and WildChat capture real-world user-LLM conversations, few studies systematically explore the mechanisms of human-LLM collaboration in coding scenarios. What tortuous paths do users experience during the interaction process? How well do the LLMs follow instructions? Are users satisfied? In this paper, we conduct an empirical analysis on human-LLM coding collaboration using LMSYS-Chat-1M and WildChat datasets to explore the human-LLM collaboration mechanism, LLMs' instruction following ability, and human satisfaction. This study yields interesting findings: 1) Task types shape interaction patterns(linear, star and tree), with code quality optimization favoring linear patterns, design-driven tasks leaning toward tree structures, and queries preferring star patterns; 2) Bug fixing and code refactoring pose greater challenges to LLMs' instruction following, with non-compliance rates notably higher than in information querying; 3) Code quality optimization and requirements-driven development tasks show lower user satisfaction, whereas structured knowledge queries and algorithm designs yield higher levels. These insights offer recommendations for improving LLM interfaces and user satisfaction in coding collaborations, while highlighting avenues for future research on adaptive dialogue systems. We believe this work broadens understanding of human-LLM synergies and supports more effective AI-assisted development.

cs.SE↗