arXiv subjects

Jian Chen

Publications and source records attributed to Jian Chen.

At least 91 records · Page 5Linked to original sources

arXiv2026

KV-CoRE: Benchmarking Data-Dependent Low-Rank Compressibility of KV-Caches in LLMs

Large language models rely on kv-caches to avoid redundant computation during autoregressive decoding, but as context length grows, reading and writing the cache can quickly saturate GPU memory bandwidth. Recent work has explored KV-cache compression, yet most approaches neglect the data-dependent nature of kv-caches and their variation across layers. We introduce KV-CoRE KV-cache Compressibility by Rank Evaluation), an SVD-based method for quantifying the data-dependent low-rank compressibility of kv-caches. KV-CoRE computes the optimal low-rank approximation under the Frobenius norm and, being gradient-free and incremental, enables efficient dataset-level, layer-wise evaluation. Using this method, we analyze multiple models and datasets spanning five English domains and sixteen languages, uncovering systematic patterns that link compressibility to model architecture, training data, and language coverage. As part of this analysis, we employ the Normalized Effective Rank as a metric of compressibility and show that it correlates strongly with performance degradation under compression. Our study establishes a principled evaluation framework and the first large-scale benchmark of kv-cache compressibility in LLMs, offering insights for dynamic, data-aware compression and data-centric model development.

cs.CL↗

arXiv2026

Q Cache: Visual Attention is Valuable in Less than Half of Decode Layers for Multimodal Large Language Model

Multimodal large language models (MLLMs) are plagued by exorbitant inference costs attributable to the profusion of visual tokens within the vision encoder. The redundant visual tokens engenders a substantial computational load and key-value (KV) cache footprint bottleneck. Existing approaches focus on token-wise optimization, leveraging diverse intricate token pruning techniques to eliminate non-crucial visual tokens. Nevertheless, these methods often unavoidably undermine the integrity of the KV cache, resulting in failures in long-text generation tasks. To this end, we conduct an in-depth investigation towards the attention mechanism of the model from a new perspective, and discern that attention within more than half of all decode layers are semantic similar. Upon this finding, we contend that the attention in certain layers can be streamlined by inheriting the attention from their preceding layers. Consequently, we propose Lazy Attention, an efficient attention mechanism that enables cross-layer sharing of similar attention patterns. It ingeniously reduces layer-wise redundant computation in attention. In Lazy Attention, we develop a novel layer-shared cache, Q Cache, tailored for MLLMs, which facilitates the reuse of queries across adjacent layers. In particular, Q Cache is lightweight and fully compatible with existing inference frameworks, including Flash Attention and KV cache. Additionally, our method is highly flexible as it is orthogonal to existing token-wise techniques and can be deployed independently or combined with token pruning approaches. Empirical evaluations on multiple benchmarks demonstrate that our method can reduce KV cache usage by over 35% and achieve 1.5x throughput improvement, while sacrificing only approximately 1% of performance on various MLLMs. Compared with SOTA token-wise methods, our technique achieves superior accuracy preservation.

cs.CV↗

arXiv2026

Arce: Augmented Roberta with Contextualized Elucidations for Ner in Automated Rule Checking

Accurate information extraction from specialized texts is a critical challenge for automated rule checking (ARC) in the architecture, engineering, and construction (AEC) domain. While large language models (LLMs) possess strong reasoning capabilities, their deployment in resource-constrained AEC environments is often impractical. Conversely, standard efficient models struggle with the significant domain gap. Although this gap can be mitigated by pre-training on large, humancurated corpora, such approaches are labor-intensive and costly. To address this, we propose ARCE (Augmented RoBERTa with Contextualized Elucidations), a novel knowledge distillation framework that leverages LLMs to synthesize a task-oriented corpus, termed Cote, for incrementally pre-training smaller models. ARCE systematically explores the optimal strategy for knowledge transfer. Our extensive experiments demonstrate that ARCE establishes a new state-of-the-art on a benchmark AEC dataset, achieving a Macro-F1 score of 77.20% and outperforming both domain-specific baselines and fine-tuned LLMs. Crucially, our study reveals a less is more principle: simple, direct explanations prove significantly more effective for domain adaptation than complex, role-based rationales in the NER task, which tend to introduce semantic noise. The source code will be made publicly available upon acceptance.

cs.CL↗

arXiv2026

Abex-rat: Synergizing Abstractive Augmentation and Adversarial Training for Classification of Occupational Accident Reports

The automatic classification of occupational accident reports is pivotal for workplace safety analysis but is persistently hindered by severe class imbalance and data scarcity. In this paper, we propose ABEX-RAT, a resource-efficient framework that synergizes generative data augmentation with robust adversarial learning. Unlike computationally expensive large language models (LLMs) fine-tuning, our approach employs a two-stage abstractive-expansive (ABEX) pipeline: it first utilizes a prompt-guided LLM to distill label-critical semantics into concise abstracts, which are then expanded into diverse synthetic samples to balance the data distribution. Subsequently, we train a lightweight classifier using a random adversarial training (RAT) protocol, which stochastically injects perturbations to enhance generalization without significant computational overhead. Experimental results on the OSHA dataset demonstrate that ABEXRAT establishes a new state-of-the-art, achieving a Macro-F1 score of 90.32% and significantly outperforming both traditional baselines and fine-tuned large models. This confirms that targeted augmentation combined with robust training offers a superior, data-efficient alternative for specialized domain classification. The source code will be made publicly available upon acceptance.

cs.LG↗

arXiv2026

MoE-Enhanced Multi-Domain Feature Selection and Fusion for Fast Map-Free Trajectory Prediction

Trajectory prediction is crucial for the reliability and safety of autonomous driving systems, yet it remains a challenging task in complex interactive scenarios due to noisy trajectory observations and intricate agent interactions. Existing methods often struggle to filter redundant scene data for discriminative information extraction, directly impairing trajectory prediction accuracy especially when handling outliers and dynamic multi-agent interactions. In response to these limitations, we present a novel map-free trajectory prediction method which adaptively eliminates redundant information and selects discriminative features across the temporal, spatial, and frequency domains, thereby enabling precise trajectory prediction in real-world driving environments. First, we design a MoE based frequency domain filter to adaptively weight distinct frequency components of observed trajectory data and suppress outlier related noise; then a selective spatiotemporal attention module that reallocates weights across temporal nodes (sequential dependencies), temporal trends (evolution patterns), and spatial nodes to extract salient information is proposed. Finally, our multimodal decoder-supervised by joint patch level and point-level losses generates reasonable and temporally consistent trajectories, and comprehensive experiments on the large-scale NuScenes and Argoverse dataset demonstrate that our method achieves competitive performance and low-latency inference performance compared with recently proposed methods.

cs.CV↗

arXiv2026

Design, waterproofing, and mass production of the 3-inch PMT frontend system of JUNO

Over 25,600 3-inch photomultiplier tubes (PMTs) have been instrumented for the central detector of the Jiangmen Underground Neutrino Observatory. Each PMT is equipped with a high-voltage divider and a frontend cable with waterproof sealing. Groups of sixteen PMTs are connected to the underwater frontend readout electronics via specialized multi-channel waterproof connectors. This paper outlines the design and mass production processes for the high-voltage divider, the cable and connector, as well as the waterproof potting of the PMT bases. The results of the acceptance tests of all the integrated PMTs are also presented.

physics.ins-det↗

arXiv2026

First-order Methods for Unconstrained Vector Optimization Problems: A Unified Majorization-Minimization Perspective

In this paper, we develop a unified majorization-minimization scheme and convergence analysis with first-order surrogate functions for unconstrained vector optimization problems (VOPs). By selecting different surrogate functions, the unified method can be reduced to various existing first-order methods. The unified convergence analysis reveals that the slow convergence of the steepest descent method is primarily attributed to the significant gap between the surrogate and objective functions. Consequently, narrowing this surrogate gap can enhance the performance of first-order methods for VOPs. To strike a better trade-off in terms of surrogate gap and per-iteration cost, we reformulate the direction-finding subproblem and elucidate that selecting a tighter surrogate function is equivalent to using an appropriate base of the dual cone in the direction-finding subproblem. Building on this insight, we employ the Barzilai-Borwein method to narrow the surrogate gap and propose a Barzilai-Borwein descent method for VOPs (BBDVO) with polyhedral cones. By reformulating the corresponding subproblem, we provide a novel perspective on the Barzilai-Borwein descent method, bridging the gap between this method and the steepest descent method. Finally, several numerical experiments are presented to validate the efficiency of the BBDVO.

math.OC↗

arXiv2026

Compliance-to-Code: Enhancing Financial Compliance Checking via Code Generation

Nowadays, regulatory compliance has become a cornerstone of corporate governance, ensuring adherence to systematic legal frameworks. At its core, financial regulations often comprise highly intricate provisions, layered logical structures, and numerous exceptions, which inevitably result in labor-intensive or comprehension challenges. To mitigate this, recent Regulatory Technology (RegTech) and Large Language Models (LLMs) have gained significant attention in automating the conversion of regulatory text into executable compliance logic. However, their performance remains suboptimal particularly when applied to Chinese-language financial regulations, due to three key limitations: (1) incomplete domain-specific knowledge representation, (2) insufficient hierarchical reasoning capabilities, and (3) failure to maintain temporal and logical coherence. One promising solution is to develop a domain specific and code-oriented datasets for model training. Existing datasets such as LexGLUE, LegalBench, and CODE-ACCORD are often English-focused, domain-mismatched, or lack fine-grained granularity for compliance code generation. To fill these gaps, we present Compliance-to-Code, the first large-scale Chinese dataset dedicated to financial regulatory compliance. Covering 1,159 annotated clauses from 361 regulations across ten categories, each clause is modularly structured with four logical elements-subject, condition, constraint, and contextual information-along with regulation relations. We provide deterministic Python code mappings, detailed code reasoning, and code explanations to facilitate automated auditing. To demonstrate utility, we present FinCheck: a pipeline for regulation structuring, code generation, and report generation.

cs.CL↗

arXiv2026

FastFLUX: Pruning FLUX with Block-wise Replacement and Sandwich Training

Recent advancements in text-to-image (T2I) generation have led to the emergence of highly expressive models such as diffusion transformers (DiTs), exemplified by FLUX. However, their massive parameter sizes lead to slow inference, high memory usage, and poor deployability. Existing acceleration methods (e.g., single-step distillation and attention pruning) often suffer from significant performance degradation and incur substantial training costs. To address these limitations, we propose FastFLUX, an architecture-level pruning framework designed to enhance the inference efficiency of FLUX. At its core is the Block-wise Replacement with Linear Layers (BRLL) method, which replaces structurally complex residual branches in ResBlocks with lightweight linear layers while preserving the original shortcut connections for stability. Furthermore, we introduce Sandwich Training (ST), a localized fine-tuning strategy that leverages LoRA to supervise neighboring blocks, mitigating performance drops caused by structural replacement. Experiments show that our FastFLUX maintains high image quality under both qualitative and quantitative evaluations, while significantly improving inference speed, even with 20\% of the hierarchy pruned. Our code will be available soon.

cs.GR↗

arXiv2026

DeKeyNLU: Enhancing Natural Language to SQL Generation through Task Decomposition and Keyword Extraction

Natural Language to SQL (NL2SQL) provides a new model-centric paradigm that simplifies database access for non-technical users by converting natural language queries into SQL commands. Recent advancements, particularly those integrating Retrieval-Augmented Generation (RAG) and Chain-of-Thought (CoT) reasoning, have made significant strides in enhancing NL2SQL performance. However, challenges such as inaccurate task decomposition and keyword extraction by LLMs remain major bottlenecks, often leading to errors in SQL generation. While existing datasets aim to mitigate these issues by fine-tuning models, they struggle with over-fragmentation of tasks and lack of domain-specific keyword annotations, limiting their effectiveness. To address these limitations, we present DeKeyNLU, a novel dataset which contains 1,500 meticulously annotated QA pairs aimed at refining task decomposition and enhancing keyword extraction precision for the RAG pipeline. Fine-tuned with DeKeyNLU, we propose DeKeySQL, a RAG-based NL2SQL pipeline that employs three distinct modules for user question understanding, entity retrieval, and generation to improve SQL generation accuracy. We benchmarked multiple model configurations within DeKeySQL RAG pipeline. Experimental results demonstrate that fine-tuning with DeKeyNLU significantly improves SQL generation accuracy on both BIRD (62.31% to 69.10%) and Spider (84.2% to 88.7%) dev datasets.

cs.AI↗

arXiv2026

ClimateIQA: A New Dataset and Benchmark to Advance Vision-Language Models in Meteorology Anomalies Analysis

Meteorological heatmaps play a vital role in deciphering extreme weather phenomena, yet their inherent complexities marked by irregular contours, unstructured patterns, and complex color variations present unique analytical hurdles for state-of-the-art Vision-Language Models (VLMs). Current state-of-the-art models like GPT-4o, Qwen-VL, and LLaVA 1.6 struggle with tasks such as precise color identification and spatial localization, resulting in inaccurate or incomplete interpretations. To address these challenges, we introduce Sparse Position and Outline Tracking (SPOT), a novel algorithm specifically designed to process irregularly shaped colored regions in visual data. SPOT identifies and localizes these regions by extracting their spatial coordinates, enabling structured representations of irregular shapes. Building on SPOT, we construct ClimateIQA, a novel meteorological visual question answering (VQA) dataset, comprising 26,280 high-resolution heatmaps and 762,120 instruction samples for wind gust, total precipitation, wind chill index and heat index analysis. ClimateIQA enhances VLM training by incorporating spatial cues, geographic metadata, and reanalysis data, improving model accuracy in interpreting and describing extreme weather features. Furthermore, we develop Climate-Zoo, a suite of fine-tuned VLMs based on SPOT-empowered ClimateIQA, which significantly outperforms existing models in meteorological heatmap tasks.

cs.CV↗

arXiv2025

Harnessing Large Language Models for Biomedical Named Entity Recognition

Background and Objective: Biomedical Named Entity Recognition (BioNER) is a foundational task in medical informatics, crucial for downstream applications like drug discovery and clinical trial matching. However, adapting general-domain Large Language Models (LLMs) to this task is often hampered by their lack of domain-specific knowledge and the performance degradation caused by low-quality training data. To address these challenges, we introduce BioSelectTune, a highly efficient, data-centric framework for fine-tuning LLMs that prioritizes data quality over quantity. Methods and Results: BioSelectTune reformulates BioNER as a structured JSON generation task and leverages our novel Hybrid Superfiltering strategy, a weak-to-strong data curation method that uses a homologous weak model to distill a compact, high-impact training dataset. Conclusions: Through extensive experiments, we demonstrate that BioSelectTune achieves state-of-the-art (SOTA) performance across multiple BioNER benchmarks. Notably, our model, trained on only 50% of the curated positive data, not only surpasses the fully-trained baseline but also outperforms powerful domain-specialized models like BioMedBERT.

cs.CL↗

arXiv2025

VividFace: High-Quality and Efficient One-Step Diffusion For Video Face Enhancement

Video Face Enhancement (VFE) aims to restore high-quality facial regions from degraded video sequences, enabling a wide range of practical applications. Despite substantial progress in the field, current methods that primarily rely on video super-resolution and generative frameworks continue to face three fundamental challenges: (1) computational inefficiency caused by iterative multi-step denoising in diffusion models; (2) faithfully modeling intricate facial textures while preserving temporal consistency; and (3) limited model generalization due to the lack of high-quality face video training data. To address these challenges, we propose VividFace, a novel and efficient one-step diffusion framework for VFE. Built upon the pretrained WANX video generation model, VividFace reformulates the traditional multi-step diffusion process as a single-step flow matching paradigm that directly maps degraded inputs to high-quality outputs with significantly reduced inference time. To enhance facial detail recovery, we introduce a Joint Latent-Pixel Face-Focused Training strategy that constructs spatiotemporally aligned facial masks to guide optimization toward critical facial regions in both latent and pixel spaces. Furthermore, we develop an MLLM-driven automated filtering pipeline that produces MLLM-Face90, a meticulously curated high-quality face video dataset, ensuring models learn from photorealistic facial textures. Extensive experiments demonstrate that VividFace achieves superior performance in perceptual quality, identity preservation, and temporal consistency across both synthetic and real-world benchmarks. We will publicly release our code, models, and dataset to support future research.

cs.CV↗

arXiv2025

What Makes a Visualization Image Complex?

We investigate the perceived visual complexity (VC) in data visualizations using objective image-based metrics. We collected VC scores through a large-scale crowdsourcing experiment involving 349 participants and 1,800 visualization images. We then examined how these scores align with 12 image-based metrics spanning information-theoretic, clutter, color, and our two object-based metrics. Our results show that both low-level image properties and the high-level elements affect perceived VC in visualization images; The number of corners and distinct colors are robust metrics across visualizations. Second, feature congestion, an information-theoretic metric capturing statistical patterns in color and texture, is the strongest predictor of perceived complexity in visualizations rich in the same stimuli; edge density effectively explains VC in node-link diagrams. Additionally, we observe a bell-curve effect for text annotations: increasing text-to-ink ratio (TiR) initially reduces complexity, reaching an optimal point, beyond which further text increases perceived complexity. Our quantification pipeline is also interpretable, enabling metric-based explanations, grounded in the VisComplexity2K dataset, bridging computational metrics with human perceptual responses. osf.io/5xe8a has the preregistration and osf.io/bdet6 has the VisComplexity2K dataset, source code, and all Apdx. and figures.

cs.HC↗

arXiv2025

Initial performance results of the JUNO detector

The Jiangmen Underground Neutrino Observatory (JUNO) started physics data taking on 26 August 2025. JUNO consists of a 20-kton liquid scintillator central detector, surrounded by a 35 kton water pool serving as a Cherenkov veto, and almost 1000 m$^2$ of plastic scintillator veto on top. The detector is located in a shallow underground laboratory with an overburden of 1800 m.w.e. This paper presents the performance results of the detector, extensively studied during the commissioning of the water phase, the subsequent liquid scintillator filling phase, and the first physics runs. The liquid scintillator achieved an attenuation length of 20.6 m at 430 nm, while the high coverage PMT system and scintillator together yielded about 1785 photoelectrons per MeV of energy deposit at the detector centre, measured using the 2.223 MeV $γ$ from neutron captures on hydrogen with an Am-C calibration source. The reconstructed energy resolution is 3.4% for two 0.511 MeV $γ$ at the detector centre and 2.9% for the 0.93 MeV quenched Po-214 alpha decays from natural radioactive sources. The energy nonlinearity is calibrated to better than 1%. Intrinsic contaminations of U-238 and Th-232 in the liquid scintillator are below 10$^{-16}$ g/g, assuming secular equilibrium. The water Cherenkov detector achieves a muon detection efficiency better than 99.9% for muons traversing the liquid scintillator volume. During the initial science runs, the data acquisition duty cycle exceeded 97.8%, demonstrating the excellent stability and readiness of JUNO for high-precision neutrino physics.

hep-ex↗

arXiv2025

First measurement of reactor neutrino oscillations at JUNO

Neutrino oscillations, a quantum effect manifesting at macroscopic scales, are governed by lepton flavor mixing angles and neutrino mass-squared differences that are fundamental parameters of particle physics, representing phenomena beyond the Standard Model. Precision measurements of these parameters are essential for testing the completeness of the three-flavor framework, determining the mass ordering of neutrinos, and probing possible new physics. The Jiangmen Underground Neutrino Observatory (JUNO) is a 20 kton liquid-scintillator detector located 52.5 km from multiple reactor cores, designed to resolve the interference pattern of reactor neutrinos with sub-percent precision. Here we report, using the first 59.1 days of data collected since detector completion in August 2025, the first simultaneous high-precision determination of two neutrino oscillation parameters, $\sin^2 θ_{12} = 0.3092\,\pm\,0.0087$ and $Δm^2_{21} = (7.50\,\pm\,0.12)\times10^{-5}\;{\rm eV}^2$ for the normal mass ordering scenario, improving the precision by a factor of 1.6 relative to the combination of all previous measurements. These results advance the basic understanding of neutrinos, validate the detector's design, and confirm JUNO's readiness for its primary goal of resolving the neutrino mass ordering with a larger dataset. The rapid achievement with a short exposure highlights JUNO's potential to push the frontiers of precision neutrino physics and paves the way for its broad scientific program.

hep-ex↗

arXiv2025

Emotion and Intention Guided Multi-Modal Learning for Sticker Response Selection

Stickers are widely used in online communication to convey emotions and implicit intentions. The Sticker Response Selection (SRS) task aims to select the most contextually appropriate sticker based on the dialogue. However, existing methods typically rely on semantic matching and model emotional and intentional cues separately, which can lead to mismatches when emotions and intentions are misaligned. To address this issue, we propose Emotion and Intention Guided Multi-Modal Learning (EIGML). This framework is the first to jointly model emotion and intention, effectively reducing the bias caused by isolated modeling and significantly improving selection accuracy. Specifically, we introduce Dual-Level Contrastive Framework to perform both intra-modality and inter-modality alignment, ensuring consistent representation of emotional and intentional features within and across modalities. In addition, we design an Intention-Emotion Guided Multi-Modal Fusion module that integrates emotional and intentional information progressively through three components: Emotion-Guided Intention Knowledge Selection, Intention-Emotion Guided Attention Fusion, and Similarity-Adjusted Matching Mechanism. This design injects rich, effective information into the model and enables a deeper understanding of the dialogue, ultimately enhancing sticker selection performance. Experimental results on two public SRS datasets show that EIGML consistently outperforms state-of-the-art baselines, achieving higher accuracy and a better understanding of emotional and intentional features. Code is provided in the supplementary materials.

cs.LG↗

arXiv2025

Revisiting Cross-Architecture Distillation: Adaptive Dual-Teacher Transfer for Lightweight Video Models

Vision Transformers (ViTs) have achieved strong performance in video action recognition, but their high computational cost limits their practicality. Lightweight CNNs are more efficient but suffer from accuracy gaps. Cross-Architecture Knowledge Distillation (CAKD) addresses this by transferring knowledge from ViTs to CNNs, yet existing methods often struggle with architectural mismatch and overlook the value of stronger homogeneous CNN teachers. To tackle these challenges, we propose a Dual-Teacher Knowledge Distillation framework that leverages both a heterogeneous ViT teacher and a homogeneous CNN teacher to collaboratively guide a lightweight CNN student. We introduce two key components: (1) Discrepancy-Aware Teacher Weighting, which dynamically fuses the predictions from ViT and CNN teachers by assigning adaptive weights based on teacher confidence and prediction discrepancy with the student, enabling more informative and effective supervision; and (2) a Structure Discrepancy-Aware Distillation strategy, where the student learns the residual features between ViT and CNN teachers via a lightweight auxiliary branch, focusing on transferable architectural differences without mimicking all of ViT's high-dimensional patterns. Extensive experiments on benchmarks including HMDB51, EPIC-KITCHENS-100, and Kinetics-400 demonstrate that our method consistently outperforms state-of-the-art distillation approaches, achieving notable performance improvements with a maximum accuracy gain of 5.95% on HMDB51.

cs.CV↗