arXiv subjects

Bowen Zhang

Publications and source records attributed to Bowen Zhang.

At least 55 records · Page 3Linked to original sources

arXiv2026

When Annotators Agree but Labels Disagree: The Projection Problem in Stance Detection

Stance detection is nearly always formulated as classifying text into Favor, Against, or Neutral. This convention was inherited from debate analysis and has been applied without modification to social media since SemEval-2016. However, attitudes toward complex targets are not unitary. A person can accept climate science while opposing carbon taxes, expressing support on one dimension and opposition on another. When annotators must compress such multi-dimensional attitudes into a single label, different annotators may weight different dimensions, producing disagreement that reflects different compression choices rather than confusion. We call this the projection problem. We conduct an annotation study across five targets from three stance benchmarks (SemEval-2016, P-Stance, COVID-19-Stance), with the same three annotators labeling all targets. For each target, annotators assign both a standard stance label and per-dimension judgments along target-specific dimensions discovered through bottom-up analysis, using the same number of categories for both. Across all fifteen target--dimension pairs, dimensional agreement consistently exceeds label agreement. The gap appears to scale with target complexity: modest for a single-entity target like Joe Biden (AC1: 0.87 vs. 0.95), but large for a multi-faceted policy target like school closures (AC1: 0.21 vs. 0.71).

cs.CL↗

arXiv2026

Filter Design for Estimating the Stellar Metallicity of Metal-poor Stars from Gaia XP Spectra

The estimation of stellar atmospheric parameters for large-scale samples, particularly metal-poor stars, is a cornerstone of Galactic archaeology. In this work, we optimized a photometric filter design tailored to measuring stellar metallicities for very metal-poor stars with [Fe/H]$< -1$.The optimal configurations consist of a central wavelength $λ_{\rm c}$ = 3960 Angstrom with a bandwidth $Δλ$ = 80 Angstrom for giant stars, and $λ_{\rm c} $= 3920 Angstrom with $Δλ$ = 80 Angstrom for dwarf stars. By applying these optimized filters to synthetic photometry derived from Gaia XP spectra, we inferred metallicities for both populations. Both internal and external validations demonstrate high precision across a wide metallicity range: 0.18-0.19 dex for $-2 \le \rm [Fe/H] \le -1$, 0.23-0.33 dex for $-3 \le \rm [Fe/H] \le -2$, and approximately 0.39 dex for the most metal-poor regime, successfully extending down to $\rm [Fe/H] \approx -4$ for giant stars, $\rm [Fe/H] \approx -3.3$ for dwarf stars. Finally, we present a catalog of approximately 14.5 million metal-poor stars with robust $\rm [Fe/H]$ measurements, along with more than ten thousand red giant ultra metal-poor candidates with $\rm [Fe/H] < -4.0$, providing a valuable resource for exploring the early formation and chemical evolution of the Milky Way.

astro-ph.SR↗

arXiv2026

RareSpot+: A Benchmark, Model, and Active Learning Framework for Small and Rare Wildlife in Aerial Imagery

Automated wildlife monitoring from aerial imagery is vital for conservation but remains limited by two persistent challenges: the difficulty of detecting small, rare species and the high cost of large-scale expert annotation. Prairie dogs exemplify this problem -- they are ecologically important yet appear tiny, sparsely distributed, and visually indistinct from their surroundings, posing a severe challenge for conventional detection models. To overcome these limitations, we present RareSpot+, a detection framework that integrates multi-scale consistency learning, context-aware augmentation, and geospatially guided active learning to address these issues. A novel multi-scale consistency loss aligns intermediate feature maps across detection heads, enhancing localization of small (approx. 30 pixels wide) objects without architectural changes, while context-aware augmentation improves robustness by synthesizing hard, ecologically plausible examples. A geospatial active learning module exploits domain-specific spatial priors linking prairie dogs and burrows, together with test-time augmentation and a meta-uncertainty model, to reduce redundant labeling. On a 2 km^2 aerial dataset, RareSpot+ improves detection over the baseline mAP@50 by +35.2% (absolute +0.13). Cross-dataset tests on HerdNet, AED, and several other wildlife benchmarks demonstrate robust detector-level transferability. The active learning module further boosts prairie dog AP by 14.5% using an annotation budget of just 1.7% of the unlabeled tiles. Beyond detection, RareSpot+ enables spatial ecological analyses such as clustering and co-occurrence, linking vision-based detection with quantitative ecology.

cs.CV↗

arXiv2026

Enhanced electron injection for efficient proton acceleration and neutron production in femtosecond laser-driven nano-structured targets

Micro- or nano-structured targets are advantageous in enhancing and manipulating laser-proton acceleration, due to the increased absorption of laser energy and onset of direct laser acceleration for high-energy electrons. Here, we experimentally demonstrate that nano-wire-array printed on a flat substrate is an efficient nano-injector of relativistic electrons that leads to a significant boost of laser-driven proton acceleration and neutron production beyond normal geometry. By employing an ultra-intense (2*1021 W/cm2) femtosecond laser pulse to irradiate nano-wire-array targets, protons with cut-off energies of 62.8 MeV are generated, and notably, the energy conversion efficiency from laser to protons reaches up to 9% - 3.5 times higher than that of flat foils. After bombarding a beryllium converter, 1.1*1010 neutrons are produced. Full 3D particle-in-cell simulations have reproduced experimental results and reveal interference mechanisms between the nano-wires and substrate, leading to continuous pumping of electrons from the substrate and standing-wave enhanced re-injection from the wire tip. This efficient injection finally results in the large sheath field and thus high yield of energetic protons and neutrons. Dependence on the wire length and scaling with laser amplitude are further discussed. These results suggest that 3D-printed structures are promising in developing compact laser-driven high-flux proton and neutron sources for numerous applications.

physics.plasm-ph↗

arXiv2026

WiFi-GEN: High-Resolution Indoor Imaging from WiFi Signals Using Generative AI

Indoor imaging is a critical task for robotics and internet-ofthings. WiFi as an omnipresent signal is a promising candidate for carrying out passive imaging and synchronizing the up-to-date information to all connected devices. This is the first research work to consider WiFi indoor imaging as a multi-modal image generation task that converts the measured WiFi power into a high-resolution indoor image. Our proposedWiFi-GEN network achieves a shape reconstruction accuracy that is 275% of that achieved by physical model-based inversion methods. Additionally, the Frechet Inception Distance score has been significantly reduced by 82%. To examine the effectiveness of models for this task, the first large-scale dataset is released containing 80,000 pairs of WiFi signal and imaging target. Our model absorbs challenges for the model-based methods including the nonlinearity, ill-posedness and non-certainty into massive parameters of our generative AI network. The network is also designed to best fit measured WiFi signals and the desired imaging output. Code: https://github.com/CNFightingSjy/WiFiGEN

cs.CV↗

arXiv2026

Leveraging Large Language Models for Semantic Query Processing in a Scholarly Knowledge Graph

The proposed research aims to develop an innovative semantic query processing system that enables users to obtain comprehensive information about research works produced by Computer Science (CS) researchers at the Australian National University (ANU). The system integrates Large Language Models (LLMs) with the ANU Scholarly Knowledge Graph (ASKG), a structured repository of all research-related artifacts produced at ANU in the CS field. Each artifact and its parts are represented as textual nodes stored in a Knowledge Graph (KG). To address the limitations of traditional scholarly KG construction and utilization methods, which often fail to capture fine-grained details, we propose a novel framework that integrates the Deep Document Model (DDM) for comprehensive document representation and the KG-enhanced Query Processing (KGQP) for optimized complex query handling. DDM enables a fine-grained representation of the hierarchical structure and semantic relationships within academic papers, while KGQP leverages the KG structure to improve query accuracy and efficiency with LLMs. By combining the ASKG with LLMs, our approach enhances knowledge utilization and natural language understanding capabilities. The proposed system employs an automatic LLM-SPARQL fusion to retrieve relevant facts and textual nodes from the ASKG. Initial experiments demonstrate that our framework is superior to baseline methods in terms of accuracy retrieval and query efficiency. We showcase the practical application of our framework in academic research scenarios, highlighting its potential to revolutionize scholarly knowledge management and discovery. This work empowers researchers to acquire and utilize knowledge from documents more effectively and provides a foundation for developing precise and reliable interactions with LLMs.

cs.IR↗

arXiv2026

The solar-like latitudinal distribution of flaring activities revealed by TESS, APOGEE and GALAH

Flare flux reflect contribution from active regions rather than the whole hemisphere of a star. Unlike the amplitude of light-curves caused by starspots, the flare detection is independent of inclination. The two valuable properties of flares can be used to reveal the latitudinal distribution of active regions (LaDAR) given that LaDAR is coupled with inclination and location information in spatially unresolved stars. We detected $\sim 27000$ flares of 1510 flaring stars in the TESS mission with the corresponding inclinations obtained. The detection rate of flaring stars shows that flares are hard to detect on stars with low inclination, indicating that flares occur mainly at low latitudes. Further investigation of the relationship between the apparent flaring activity and inclination along with the rotation period finds that as the rotation period increases from a solar-like rotation to an ultra-fast rotation, the mean latitude of active regions increases from $θ\approx 15^{\circ}$ to $θ\approx 27^{\circ}$, whose trend is in line with the rotation--activity relationship. The LaDAR indicates that flares are attributed to small-scale fields that are formed at low latitudes, while polar spots that are associated with large-scale fields are inactive and are difficult to trigger flares.

astro-ph.SR↗

arXiv2026

Stress-driven dynamic evolution of core-shell structured cavities with H and He in BCC-Fe under fusion conditions

Understanding the dynamic behavior of microstructures formed under fusion conditions is critical for designing high-performance structural materials for fusion reactors. Under fusion conditions, cavities of core-shell structures are formed due to the interaction between irradiation-induced vacancies and H and He atoms produced via transmutation. In this study, thermodynamic analysis and molecular dynamics simulations are combined to investigate the atomic-scale mechanisms and dynamic response of core-shell cavities formed in BCC-Fe under applied stress/strain fields. The thermodynamic analysis provides both the foundational reference for cavity structures under fusion neutron irradiation and the initial configurations for atomistic simulations. Building on this framework, atomic-scale simulations demonstrate that H and He play a decisive role in the stress-strain response and the evolution of elastic-plastic deformation within the cavities. In core-shell configurations, H atoms serve a function analogous to that in He-filled cavities, synergistically interacting with He to induce cavity deformation under mechanical loading.

cond-mat.mtrl-sci↗

arXiv2026

PoLi-RL: A Point-to-List Reinforcement Learning Framework for Conditional Semantic Textual Similarity

Conditional Semantic Textual Similarity (C-STS) measures the semantic proximity between text segments under a specific condition, thereby overcoming the ambiguity inherent in traditional STS. However, existing methods are largely confined to discriminative models, failing to fully leverage recent breakthroughs in the NLP community involving Large Language Models (LLMs) and Reinforcement Learning (RL). RL is a particularly well-suited paradigm for this task, as it can directly optimize the non-differentiable Spearman ranking metric and guide the reasoning process required by C-STS. Nevertheless, we find that naively applying listwise RL fails to produce meaningful improvements, as the model struggles with complex, coarse-grained reward signals, leading to optimization difficulties. To address this challenge, we introduce PoLi-RL, a novel Point-to-List Reinforcement Learning framework. PoLi-RL employs a two-stage curriculum: it first trains the model with a simple pointwise reward to establish fundamental scoring capabilities, then transitions to a hybrid reward that combines pointwise, pairwise, and listwise objectives to refine the model's ability to discern subtle semantic distinctions. Crucially, we propose an innovative Parallel Slice Ranking Reward (PSRR) mechanism that computes ranking rewards in parallel slices, where each slice consists of completions with the same index from different samples. This provides a precise, differentiated learning signal for each individual completion, enabling granular credit assignment and effective optimization. On the official C-STS benchmark, PoLi-RL achieves a Spearman correlation coefficient of 48.18, establishing a new SOTA for the cross-encoder architecture. As the first work to successfully apply RL to C-STS, our study introduces a powerful paradigm for aligning LLMs for complex, ranking-based conditional judgment tasks.

cs.CL↗

arXiv2026

MiniCPM-SALA: Hybridizing Sparse and Linear Attention for Efficient Long-Context Modeling

The evolution of large language models (LLMs) towards applications with ultra-long contexts faces challenges posed by the high computational and memory costs of the Transformer architecture. While existing sparse and linear attention mechanisms attempt to mitigate these issues, they typically involve a trade-off between memory efficiency and model performance. This paper introduces MiniCPM-SALA, a 9B-parameter hybrid architecture that integrates the high-fidelity long-context modeling of sparse attention (InfLLM-V2) with the global efficiency of linear attention (Lightning Attention). By employing a layer selection algorithm to integrate these mechanisms in a 1:3 ratio and utilizing a hybrid positional encoding (HyPE), the model maintains efficiency and performance for long-context tasks. Furthermore, we introduce a cost-effective continual training framework that transforms pre-trained Transformer-based models into hybrid models, which reduces training costs by approximately 75% compared to training from scratch. Extensive experiments show that MiniCPM-SALA maintains general capabilities comparable to full-attention models while offering improved efficiency. On a single NVIDIA A6000D GPU, the model achieves up to 3.5x the inference speed of the full-attention model at the sequence length of 256K tokens and supports context lengths of up to 1M tokens, a scale where traditional full-attention 8B models fail because of memory constraints.

cs.CL↗

arXiv2026

AXLearn: Modular, Hardware-Agnostic Large Model Training

AXLearn is a production system which facilitates scalable and high-performance training of large deep learning models. Compared to other state-of-art deep learning systems, AXLearn has a unique focus on modularity and support for hardware-agnostic training. AXLearn's internal interfaces between software components follow strict encapsulation, allowing different components to be assembled to facilitate rapid model development and experimentation on different hardware infrastructure. AXLearn maintains constant complexity as we scale the components in the system, compared to linear or quadratic complexity in state-of-the-art training systems. This allows integrating features such as Rotary Position Embeddings (RoPE) into AXLearn across hundred of modules with just 10 lines of code, compared to hundreds as required in other systems. At the same time, AXLearn maintains equivalent performance compared to state-of-the-art training systems. Finally, we share our experience in the development and operation of AXLearn at Apple.

cs.LG↗

arXiv2026

CLARITY: Contextual Linguistic Adaptation and Accent Retrieval for Dual-Bias Mitigation in Text-to-Speech Generation

Instruction-guided text-to-speech (TTS) research has reached a maturity level where excellent speech generation quality is possible on demand, yet two coupled biases persist in reducing perceived quality: accent bias, where models default towards dominant phonetic patterns, and linguistic bias, a misalignment in dialect-specific lexical or cultural information. These biases are interdependent and authentic accent generation requires both accent fidelity and correctly localized text. We present CLARITY (Contextual Linguistic Adaptation and Retrieval for Inclusive TTS sYnthesis), a backbone-agnostic framework to address both biases through dual-signal optimization. Firstly, we apply contextual linguistic adaptation to localize input text to align with the target dialect. Secondly, we propose retrieval-augmented accent prompting (RAAP) to ensure accent-consistent speech prompts. We evaluate CLARITY on twelve varieties of English accent via both subjective and objective analysis. Results clearly indicate that CLARITY improves accent accuracy and fairness, ensuring higher perceptual quality output\footnote{Code and audio samples are available at https://github.com/ICT-SIT/CLARITY.

cs.SD↗

arXiv2026

Cross-Lingual F5-TTS: Towards Language-Agnostic Voice Cloning and Speech Synthesis

Flow-matching-based text-to-speech (TTS) models have shown high-quality speech synthesis. However, most current flow-matching-based TTS models still rely on reference transcripts corresponding to the audio prompt for synthesis. This dependency prevents cross-lingual voice cloning when audio prompt transcripts are unavailable, particularly for unseen languages. The key challenges for flow-matching-based TTS models to remove audio prompt transcripts are identifying word boundaries during training and determining appropriate duration during inference. In this paper, we introduce Cross-Lingual F5-TTS, a framework that enables cross-lingual voice cloning without audio prompt transcripts. Our method preprocesses audio prompts by forced alignment to obtain word boundaries, enabling direct synthesis from audio prompts while excluding transcripts during training. To address the duration modeling challenge, we train speaking rate predictors at different linguistic granularities to derive duration from speaker pace. Experiments show that our approach matches the performance of F5-TTS while enabling cross-lingual voice cloning.

cs.SD↗

arXiv2026

Not All Layers Need Tuning: Selective Layer Restoration Recovers Diversity

Post-training improves instruction-following and helpfulness of large language models (LLMs) but often reduces generation diversity, which leads to repetitive outputs in open-ended settings, a phenomenon known as mode collapse. Motivated by evidence that LLM layers play distinct functional roles, we hypothesize that mode collapse can be localized to specific layers and that restoring a carefully chosen range of layers to their pre-trained weights can recover diversity while maintaining high output quality. To validate this hypothesis and decide which layers to restore, we design a proxy task -- Constrained Random Character(CRC) -- with an explicit validity set and a natural diversity objective. Results on CRC reveal a clear diversity-validity trade-off across restoration ranges and identify configurations that increase diversity with minimal quality loss. Based on these findings, we propose Selective Layer Restoration (SLR), a training-free method that restores selected layers in a post-trained model to their pre-trained weights, yielding a hybrid model with the same architecture and parameter count, incurring no additional inference cost. Across three different tasks (creative writing, open-ended question answering, and multi-step reasoning) and three different model families (Llama, Qwen, and Gemma), we find SLR can consistently and substantially improve output diversity while maintaining high output quality.

cs.CL↗

arXiv2026

Physical properties of RhGe and CoGe single crystals synthesized under high pressure

Chiral topological semimetals hosting multifold fermions and exotic surface states represent a frontier in topological materials research. Among them, noncentrosymmetric cubic B20 compounds-notably transition-metal silicides and germanides-offer a unique platform for realizing symmetry-protected topological phases and unconventional optoelectronic responses. Here, we report the physical properties of RhGe and CoGe single crystals with B20 structure in detail. Transport measurements reveal metallic behavior with characteristic Fermi-liquid scaling at low temperatures, while magnetization results confirm paramagnetism in both compounds. In addition, both of materials exhibit low carrier concentrations with small electronic specific heat coefficient, indicating their semimetal feature with weak electronic correlations. Such high-quality CoGe and RhGe single crystals provide a material platform to explore the evolution of multifold fermions and the instability of helicoid-arc surface states with spin-orbit coupling and surface environment in B20 material systems.

cond-mat.str-el↗

arXiv2026

Audio ControlNet for Fine-Grained Audio Generation and Editing

We study the fine-grained text-to-audio (T2A) generation task. While recent models can synthesize high-quality audio from text descriptions, they often lack precise control over attributes such as loudness, pitch, and sound events. Unlike prior approaches that retrain models for specific control types, we propose to train ControlNet models on top of pre-trained T2A backbones to achieve controllable generation over loudness, pitch, and event roll. We introduce two designs, T2A-ControlNet and T2A-Adapter, and show that the T2A-Adapter model offers a more efficient structure with strong control ability. With only 38M additional parameters, T2A-Adapter achieves state-of-the-art performance on the AudioSet-Strong in both event-level and segment-level F1 scores. We further extend this framework to audio editing, proposing T2A-Editor for removing and inserting audio events at time locations specified by instructions. Models, code, dataset pipelines, and benchmarks will be released to support future research on controllable audio generation and editing.

cs.SD↗

arXiv2026

ERNIE 5.0 Technical Report

In this report, we introduce ERNIE 5.0, a natively autoregressive foundation model desinged for unified multimodal understanding and generation across text, image, video, and audio. All modalities are trained from scratch under a unified next-group-of-tokens prediction objective, based on an ultra-sparse mixture-of-experts (MoE) architecture with modality-agnostic expert routing. To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. Within a single pre-training run, the model learns a family of sub-models with varying depths, expert capacities, and routing sparsity, enabling flexible trade-offs among performance, model size, and inference latency in memory- or time-constrained scenarios. Moreover, we systematically address the challenges of scaling reinforcement learning to unified foundation models, thereby guaranteeing efficient and stable post-training under ultra-sparse MoE architectures and diverse multimodal settings. Extensive experiments demonstrate that ERNIE 5.0 achieves strong and balanced performance across multiple modalities. To the best of our knowledge, among publicly disclosed models, ERNIE 5.0 represents the first production-scale realization of a trillion-parameter unified autoregressive model that supports both multimodal understanding and generation. To facilitate further research, we present detailed visualizations of modality-agnostic expert routing in the unified model, alongside comprehensive empirical analysis of elastic training, aiming to offer profound insights to the community.

cs.CL↗

arXiv2026

HY3D-Bench: Generation of 3D Assets

While recent advances in neural representations and generative models have revolutionized 3D content creation, the field remains constrained by significant data processing bottlenecks. To address this, we introduce HY3D-Bench, an open-source ecosystem designed to establish a unified, high-quality foundation for 3D generation. Our contributions are threefold: (1) We curate a library of 250k high-fidelity 3D objects distilled from large-scale repositories, employing a rigorous pipeline to deliver training-ready artifacts, including watertight meshes and multi-view renderings; (2) We introduce structured part-level decomposition, providing the granularity essential for fine-grained perception and controllable editing; and (3) We bridge real-world distribution gaps via a scalable AIGC synthesis pipeline, contributing 125k synthetic assets to enhance diversity in long-tail categories. Validated empirically through the training of Hunyuan3D-2.1-Small, HY3D-Bench democratizes access to robust data resources, aiming to catalyze innovation across 3D perception, robotics, and digital content creation.

cs.CV↗