arXiv subjects

Zhu Liu

Publications and source records attributed to Zhu Liu.

At least 19 recordsLinked to original sources

arXiv2026

Machine learning emulators of dynamic global vegetation models for fast gridded carbon budget predictions

The land carbon sink explains most of the variability of the atmospheric CO2 growth rate, but the Global Carbon Budget estimates it with dynamic global vegetation models (DGVMs) run once a year. Emulators could provide it sooner if they reproduce each model outside its training period. Here we show that convolutional neural network emulators of 19 of these models, driven by ERA5 climate and their own past predictions, reproduce the global land sink of these models and its spread over two years not used in training. Global monthly net biome production (NBP) anomalies were reproduced with a median R^2 of 0.81. The error of the multi-model mean annual NBP was at most 0.23 GtC yr^-1. At the grid-cell level, the median R^2 of monthly anomalies was 0.30, partly because of errors that accumulate in the emulated NBP history.

physics.ao-ph↗

arXiv2026

CTFS : Collaborative Teacher Framework for Forward-Looking Sonar Image Semantic Segmentation with Extremely Limited Labels

As one of the most important underwater sensing technologies, forward-looking sonar exhibits unique imaging characteristics. Sonar images are often affected by severe speckle noise, low texture contrast, acoustic shadows, and geometric distortions. These factors make it difficult for traditional teacher-student frameworks to achieve satisfactory performance in sonar semantic segmentation tasks under extremely limited labeled data conditions. To address this issue, we propose a Collaborative Teacher Semantic Segmentation Framework for forward-looking sonar images. This framework introduces a multi-teacher collaborative mechanism composed of one general teacher and multiple sonar-specific teachers. By adopting a multi-teacher alternating guidance strategy, the student model can learn general semantic representations while simultaneously capturing the unique characteristics of sonar images, thereby achieving more comprehensive and robust feature modeling. Considering the challenges of sonar images, which can lead teachers to generate a large number of noisy pseudo-labels, we further design a cross-teacher reliability assessment mechanism. This mechanism dynamically quantifies the reliability of pseudo-labels by evaluating the consistency and stability of predictions across multiple views and multiple teachers, thereby mitigating the negative impact caused by noisy pseudo-labels. Notably, on the FLSMD dataset, when only 2% of the data is labeled, our method achieves a 5.08% improvement in mIoU compared to other state-of-the-art approaches.

cs.CV↗

arXiv2026

Behave Like Experts: Fusing Data with Semantic Knowledge for Effective and Efficient Semantic Map Construction

Semantic map models (SMMs) construct a graph that reflects conceptual similarity by comparing cross-lingual form-meaning mappings. Experts build such models not only by obeying the data constraint, i.e., the connectivity hypothesis, which states that conceptual nodes belonging to the same form must be connected, but also by incorporating their prior semantic knowledge. Current automatic methods ignore this knowledge, which harms both accuracy and efficiency by enlarging the search space of candidate graphs. In this paper, we fuse the distributional semantic knowledge hidden in the internal representations of language models into a top-down, data-driven algorithm. This expert-like behavior re-sorts the candidates in a huge search space, guided by conceptual semantics similarity. Experiments on a self-collected benchmark of 10 datasets and comprehensive metrics show that our method builds SMMs both effectively and efficiently.

cs.CL↗

arXiv2026

Retrieve What Helps, Only When It Helps: Agentic Reinforcement Learning for Selective Retrieval-Augmented Machine Translation

Retrieving external knowledge can improve large language model (LLM) translation by providing supplementary information such as idiom interpretations and terminology. However, existing retrieval-augmented machine translation systems typically apply predefined retrieval strategies uniformly across all inputs. We argue that translation operates in a partial-evidence regime: LLM's parametric knowledge suffices for many inputs, while external evidence benefits a subset and usually supports rather than directly provides the desired translation. This creates two challenges: deciding when to retrieve and what evidence to acquire. We introduce Selective Retrieval-Augmented Translation (SRAT), an agentic reinforcement learning framework that learns when to translate directly or retrieve, what to retrieve, and how to integrate these decisions by optimizing final translation quality. To align evidence retrieval with downstream translation quality rather than simple semantic similarity, we further develop joint Plackett--Luce optimization and translation-quality contrastive training for retriever-training. Experiments on idiom, knowledge-intensive, and document-level translation with offline or online retrieval show that SRAT consistently outperforms never- and always-retrieve baselines, and quality-aligned retriever training provides further gains when similarity-based retrieval is insufficient. The results demonstrate the benefit of treating retrieval as a learnable translation action rather than fixed preprocessing.

cs.CL↗

arXiv2026

IRSTD-Agent: Agentic Infrared Small Target Detection via Zoom-Guided Interaction Learning

Infrared small-target detection plays an important role in maritime monitoring and aerial surveillance. Although multimodal large language models (MLLMs) offer promising capabilities for visual understanding, existing MLLM-based approaches struggle to precisely localize infrared small targets. In this paper, we propose IRSTD-Agent, an agentic framework for infrared small target detection through dynamic visual search. The framework enables an MLLM to adaptively determine where and at what scale to inspect an image and progressively gather fine-grained visual evidence for precise target localization. Five complementary visual tools (PROPOSAL, ZOOM, DETECT, DROP and REFINE) support object candidate discovery, adaptive observation, target localization, hypothesis rejection, and target extent refinement, together enabling a coordinated search process over original-resolution images. To teach the MLLMs to conduct this search, we introduce Zoom-guided Interaction Learning, which uses annotation-derived interaction trajectories to supervise tool selection and the corresponding arguments. Through extensive experiments on WideIRSTD-Full and IRSTD-1k datasets, we demonstrate that IRSTD-Agent outperforms the evaluated vision-language models and enhances the precise localization capabilities of MLLMs in IRSTD tasks.

cs.CV↗

arXiv2026

Breaking Babel: A Self-Evolving Multi-Agent System for Long-Form Subtitle Translation

Long-form subtitle translation requires reasoning over discourse and cultural context spanning episodes or entire series, while maintaining consistent terminology and style. Existing single-LLM methods are largely sentence-level, and multi-agent systems often use static workflows that do not adapt to scene complexity or production context. We propose SMART, a Self-evolving Multi-Agent system for long-foRm subtitle Translation. During test-time training, SMART builds persistent series-level memory and translates a subset of sentences through a dynamic router and Mixture-of-Agents layer with tools for terminology verification, subtitle constraint validation, and contextual retrieval. A judge-refiner loop scores candidates and uses textual critiques to update agent prompts and routing policies without retraining the underlying LLMs. During test-time inference, the evolved configuration translates the remaining series. We also introduce Subtitle Arena, covering 14 genres, 2--198 episodes per series, production years 1959--2023, and 15 target locales, together with SubMQM, a subtitle-adapted MQM framework with seven dimensions and 19 error categories. SMART achieves the best overall MQM score in all 15 Subtitle Arena directions, reducing average penalty by 6.9% over the strongest competing agent system. On a public benchmark, MuSC, SMART obtains the best model result across all 4 language pairs. SMART also achieves the best result in human evaluation with an overall score of 4.50/5.

cs.CL↗

arXiv2026

Low latency global carbon budget reveals strong land sink recovery in 2025

The atmospheric CO2 growth rate fell sharply in 2025, from a record 3.76 $\pm$ 0.09 ppm yr-1 in 2024 to 2.06 $\pm$ 0.09 ppm yr-1 (NOAA marine boundary layer observations), below the 2015-2022 mean of 2.47 ppm yr-1, even as fossil CO2 emissions rose by 0.7% to 10.38 GtC yr-1. Here we present a low-latency global and regional carbon budget for 2025, combining three dynamic global vegetation models (DGVMs) and ocean model emulators with four atmospheric inversions constrained by OCO-2 satellite retrievals. The global net land sink reached 2.36 $\pm$ 0.16 GtC yr-1 in 2025 (DGVMs: 2.04 $\pm$ 0.24; inversions: 2.68 $\pm$ 0.20 GtC yr-1), strengthening by 2.81 $\pm$ 0.31 GtC yr-1 from 2024 and exceeding the 2015-2022 mean by 0.71 $\pm$ 0.13 GtC yr-1. Ocean uptake (3.11 $\pm$ 0.36 GtC yr-1) remained similar to 2024, making the land sink rebound the dominant driver of the slowdown in CO2 growth. Tropical lands shifted from net sources in 2024 to net sinks in 2025, with enhanced uptake across much of Africa and northern Eurasia, and land flux anomalies covaried with GRACE terrestrial water storage. Where the sink had weakened substantially in 2023-2024, about 80% of the area showed some recovery, with overall recovery of 87.3% (DGVMs) to 99.5% (inversions). Recovery exceeded 100% in the tropics but remained incomplete in the northern extratropics, indicating a strong but spatially uneven rebound of the land carbon sink.

physics.ao-ph↗

arXiv2026

DISTA-Net++: Rethinking Infrared Small Target Unmixing Beyond Sub-Pixel Separation

Long-range infrared imaging frequently confronts dense target clusters whose diffraction-limited signatures merge into a single indistinguishable blob, concealing the number, sub-pixel positions, and radiant intensities of the underlying sources. While deep learning has advanced general object detection, resolving such Closely-Spaced Infrared Small Targets (CSIST) remains largely unexplored, owing to a systemic infrastructure void and a fundamental paradigm mismatch. The dominant formulation, which reduces unmixing to a blind, discrete sub-pixel separation, is inherently insufficient: without semantic guidance, the ill-posed inverse problem admits ambiguous solutions plagued by false and missed detections, while grid-based discretization locks predictions onto fixed lattice centers, chaining precision to prohibitively expensive grid refinement. We argue that CSIST unmixing should instead be informed and continuous. To ground this paradigm shift, we establish the first comprehensive open-source ecosystem for the field, comprising the large-scale CSIST-100K benchmark, a tailored metric suite, and the GrokCSO toolkit. Upon this foundation, we propose DISTA-Net++, which anchors a dynamic deep unfolding backbone with two synergistic mechanisms: a Count-Guided Prior that injects the global target count as an explicit semantic constraint to regularize the solution space, and a Continuous Coordinate Rectification that regresses off-grid offsets to decouple localization accuracy from grid resolution. Extensive experiments validate our paradigm: even under the most economical 3x division, DISTA-Net++ surpasses 7x-division state-of-the-art methods by 16.15% in CSO-mAP and 62.96% in count accuracy at merely one-sixth of their computation, demonstrating that unmixing precision need not be purchased with finer discretization. The complete ecosystem is available at https://github.com/GrokCV/GrokDet.

cs.CV↗

arXiv2026

CtrlSpeech: Coarse-to-Fine Control for Expressive Speech Synthesis

Recent Text-To-Speech (TTS) systems have achieved strong naturalness and zero-shot voice cloning performance, but fine-grained control of expressive speech at the word or phoneme level remains challenging. We propose CtrlSpeech, a controllable, expressive TTS framework with coarse-to-fine control. Built on the DiTAR architecture, CtrlSpeech combines global speaker conditioning with phone-aligned pitch, loudness, and duration signals, enabling localized prosodic control while preserving the target speaker's timbre. This design allows users to adjust expressive attributes at a fine temporal granularity, making speech refinement more flexible and controllable. Experimental results show that CtrlSpeech achieves competitive zero-shot TTS performance and improves controllability over expressive attributes, demonstrating its effectiveness for flexible and practical expressive speech synthesis.

eess.AS↗

arXiv2026

AugCodec: A Low-Bitrate Disentangled Neural Speech Codec via Data Augmentation

We propose AugCodec, a low-bitrate disentangled neural speech codec that leverages data augmentation to decompose speech into three distinct components: semantic, speaker, and prosody tokens. Specifically, we employ tailored augmenta tion strategies to transform speech into distinct variants, each serving as input for extracting tokens that preserve the target attribute while suppressing others. This disentanglement strategy enables substantial reduction in token rate. Further more, we introduce an augmentation loss that aligns semantic encoder outputs between source and voice-converted speech, encouraging speaker-agnostic embeddings while mitigating the acoustic mismatch induced by voice conversion. Experiments on LibriSpeech test-clean demonstrate that AugCodec significantly outperforms state-of-the-art methods in both reconstruction quality and disentanglement, while operating at only 12.5Hz with three token streams.

cs.SD↗

arXiv2026

Unveiling Orbital-mediated Ultrafast Demagnetization in Rare Earth-Transition-Metal Ferrimagnets

The ultimate speed limit of magnetic recording and spintronic devices is set by the efficiency of angular-momentum transfer during ultrafast demagnetization, yet its microscopic pathway in Rare-Earth-Transition-Metal (RE-TM) ferrimagnets remains debated. Here, we establish an orbital-mediated framework in which 3d spin-orbit coupling (SOC) governs angular momentum (AM) dissipation. Strong 3d-SOC in RE-Co enables sub-picosecond, single-step demagnetization via direct orbital-to-lattice transfer, whereas weak 3d-SOC in RE-Fe redirects AM into 4f orbitals, producing slower two-step dynamics. The second-stage rate scales with 4f-SOC strength, revealing a distinct orbital-mediated dissipation channel. Using time-resolved magneto-optical Kerr measurements, supported by an extended four-temperature model, corroborate this picture across diverse RE-TM systems (RE = Sm, Gd, Tb, Dy, Ho and TM = Fe, Co, CoNi). Our results identify the SOC-driven competition between 3d and 4f orbital channels as the universal mechanism governing ultrafast demagnetization in RE-TM ferrimagnets, enabling rational design of the switching speed for next-generation spintronic devices.

cond-mat.mtrl-sci↗

arXiv2026

Diffuse to Detect: Bi-Level Sample Rebalancing with Pseudo-Label Diffusion for Point-Supervised Infrared Small-Target Detection

Point supervision has become a scalable solution to address dense annotation for infrared small target detection, but its performance is limited by two coupled bottlenecks: unstable pseudo-label evolution in cluttered, low-contrast infrared imagery and severe sample-distribution imbalance. In this paper, we present a more adaptive and stable framework to address these issues. Leveraging the intrinsic consistency between thermal radiation patterns and heat diffusion, we propose a physics-induced annotation strategy that expands single-point labels into reliable pseudo-masks. To further enhance supervision and alleviate sample imbalance, we develop a bi-level dual-update framework that jointly optimizes detector weights, sample weights, and diffusion parameters. A meta-classifier dynamically predicts sample-wise loss weights, while a differentiable diffusion module refines pseudo-labels with detection feedback, enabling adaptive interaction between training and hyperparameter optimization. Extensive experiments across multiple datasets demonstrate five-fold annotation acceleration, superior detection accuracy, and comparable performance with 30% of the training data, validating the efficiency and practicality of our approach. Our code is available at https://github.com/yuanhang-yao/diffuse-to-detect.

cs.CV↗

arXiv2026

Learning with Semantic Priors: Stabilizing Point-Supervised Infrared Small Target Detection via Hierarchical Knowledge Distillation

Single-frame Infrared Small Target Detection (ISTD) aims to localize weak targets under heavy background clutter, yet dense pixel-wise annotations are expensive. Point supervision with online label evolution reduces annotation cost; however, lightweight CNN detectors often lack sufficient semantics, leading to noisy pseudo-masks and unstable optimization. To address this, we propose a hierarchical VFM-driven knowledge distillation framework that uses a frozen Vision Foundation Model (VFM) during training. We formulate point-supervised learning as a bilevel optimization process: the inner loop adapts a VFM-embedded teacher on reweighted training samples, while the outer loop transfers validation-guided knowledge to a lightweight student to mitigate pseudo-label noise and training-set bias. We further introduce Semantic-Conditioned Affine Modulation (SCAM) to inject VFM semantics into CNN features at multiple layers. In addition, a dynamic collaborative learning strategy with cluster-level sample reweighting enhances robustness to imperfect pseudo-masks. Experiments on diverse challenging cases across multiple ISTD backbones demonstrate consistent improvements in detection accuracy and training stability. Our code is available at https://github.com/yuanhang-yao/semantic-prior.

cs.CV↗

arXiv2026

ImCoref-CeS: An Improved Lightweight Pipeline for Coreference Resolution with LLM-based Checker-Splitter Refinement

Coreference Resolution (CR) is a critical task in Natural Language Processing (NLP). Current research faces a key dilemma: whether to further explore the potential of supervised neural methods based on small language models, whose detect-then-cluster pipeline still delivers top performance, or embrace the powerful capabilities of Large Language Models (LLMs). However, effectively combining their strengths remains underexplored. To this end, we propose \textbf{ImCoref-CeS}, a novel framework that integrates an enhanced supervised model with LLM-based reasoning. First, we present an improved CR method (\textbf{ImCoref}) to push the performance boundaries of the supervised neural method by introducing a lightweight bridging module to enhance long-text encoding capability, devising a biaffine scorer to comprehensively capture positional information, and invoking a hybrid mention regularization to improve training efficiency. Importantly, we employ an LLM acting as a multi-role Checker-Splitter agent to validate candidate mentions (filtering out invalid ones) and coreference results (splitting erroneous clusters) predicted by ImCoref. Extensive experiments demonstrate the effectiveness of ImCoref-CeS, which achieves superior performance compared to existing state-of-the-art (SOTA) methods.

cs.CL↗

arXiv2026

Labeled TrustSet Guided: Batch Active Learning with Reinforcement Learning

Batch active learning (BAL) is a crucial technique for reducing labeling costs and improving data efficiency in training large-scale deep learning models. Traditional BAL methods often rely on metrics like Mahalanobis Distance to balance uncertainty and diversity when selecting data for annotation. However, these methods predominantly focus on the distribution of unlabeled data and fail to leverage feedback from labeled data or the model's performance. To address these limitations, we introduce TrustSet, a novel approach that selects the most informative data from the labeled dataset, ensuring a balanced class distribution to mitigate the long-tail problem. Unlike CoreSet, which focuses on maintaining the overall data distribution, TrustSet optimizes the model's performance by pruning redundant data and using label information to refine the selection process. To extend the benefits of TrustSet to the unlabeled pool, we propose a reinforcement learning (RL)-based sampling policy that approximates the selection of high-quality TrustSet candidates from the unlabeled data. Combining TrustSet and RL, we introduce the Batch Reinforcement Active Learning with TrustSet (BRAL-T) framework. BRAL-T achieves state-of-the-art results across 10 image classification benchmarks and 2 active fine-tuning tasks, demonstrating its effectiveness and efficiency in various domains.

cs.LG↗

arXiv2026

Global near-real-time daily emissions of atmospheric pollutants from power plants

The power sector is a major source of fossil fuel use and air pollutant emissions, making high-spatiotemporal-resolution emission accounting essential for effective mitigation policy and air quality management. Yet existing public inventories are often limited by low timeliness and coarse resolution. Here, we develop a global, plant-level, daily, multi-pollutant emission database for the power sector by integrating nearly 3 million hourly-to-daily near-real-time power generation records from 57 countries, representing about 81% of global fossil-fuel-based electricity generation, with fundamental information for more than 10,000 power plants worldwide, including location and installed capacity. The dataset substantially improves the timeliness and granularity of global power-sector emission estimates. From 2019 to 2025, emissions of most pollutants increased, with 2025 daily mean emissions reaching 0.274 kt/d for BC, 45.1 kt/d for CO, 0.418 kt/d for NH3, 52.2 kt/d for NOx, 3.01 kt/d for NMVOC, 0.418 kt/d for OC, 6.76 kt/d for PM10, 5.11 kt/d for PM2.5, and 78.5 kt/d for SO2. Compared with 2019, NMVOC showed the largest increase, whereas SO2 was the only pollutant to decline overall. Coal remained the dominant source of sulfur-, nitrogen-, and particulate-related emissions, while gas and biomass contributed more to carbonaceous species and reduced nitrogen. The dataset also captures pronounced seasonal, regional, and short-term variability. Against EDGAR for 2019-2022, our estimates agree well, with Pearson correlations of 0.92-0.99 and mean relative deviations of 8.8%-28.1%. This near-real-time, high-resolution dataset provides a strong foundation for air pollution control, carbon mitigation, emission monitoring, and satellite-based inversion.

physics.ao-ph↗

arXiv2026

Near real-time monitoring of global land-ocean cover dynamics

Monitoring the dynamics of global land-ocean cover is fundamental for regulating the Earth's climate and sustaining terrestrial and marine ecosystems. However, existing datasets and research often exhibit limitations in temporal resolution and timeliness, lack coupled analysis of land cover and sea ice dynamics, and fail to incorporate the perspective of Earth system safety thresholds. Here, we developed an integrated monitoring framework by fusing multi-source remote sensing and reanalysis data, generating a 5-day resolution time series (2018-2025) of global land cover and sea ice coverage with near-real-time update capability. Our analysis reveals distinct latitudinal and regional patterns, with forests dominating (27.0% of global land area) tropical and subtropical regions. At the national scale, land cover composition and seasonal rhythms vary significantly, with countries like China, India, and the US exhibiting divergent patterns such as bimodal cropland fluctuations and alternating snow/ice dominance. Temporally, vegetated cover types exhibit seasonal cycles peaking during Northern Hemisphere summer, and a pronounced anti-phase seasonal pattern is observed between Arctic and Antarctic sea ice coverage. Crucially, safety threshold analysis indicates the global forest cover indicator (~60%) is approaching the 54% lower safe limit, with a declining trend in recent years. Concurrently, Arctic sea ice coverage in September occasionally drops to 23%, below its critical upper limit of 27.6%. Temperature presents a significant negative correlation with sea ice cover (R = -0.78, p < 0.001), with asymmetric freezing and melting rates. By quantifying the proximity of key indicators to their safety thresholds, this study provides a robust, integrated framework for early-warning assessment, thereby offering vital scientific support for global climate adaptation and sustainable policymaking.

physics.ao-ph↗

arXiv2026

XISM: an eXploratory and Interactive Graph Tool to Visualize and Evaluate Semantic Map Models

Semantic map models visualize systematic relations among semantic functions through graph structures and are widely used in linguistic typology. However, existing construction methods either depend on labor-intensive expert reasoning or on fully automated systems lacking expert involvement, creating a tension between scalability and interpretability. We introduce \textbf{XISM}, an interactive system that combines data-driven inference with expert knowledge. XISM generates candidate maps via a top-down procedure and allows users to iteratively refine edges in a visual interface, with real-time metric feedback. Experiments in three semantic domains and expert interviews show that XISM improves linguistic decision transparency and controllability in semantic-map construction while maintaining computational efficiency. XISM provides a collaborative approach for scalable and interpretable semantic-map building. The system\footnote{https://app.xism2025.xin/} , source code\footnote{https://github.com/hank317/XISM} , and demonstration video\footnote{https://youtu.be/m5laLhGn6Ys} are publicly available.

cs.CL↗