arXiv subjects

Chuan Li

Publications and source records attributed to Chuan Li.

At least 73 records · Page 4Linked to original sources

arXiv2025

JAM: A Tiny Flow-based Song Generator with Fine-grained Controllability and Aesthetic Alignment

Diffusion and flow-matching models have revolutionized automatic text-to-audio generation in recent times. These models are increasingly capable of generating high quality and faithful audio outputs capturing to speech and acoustic events. However, there is still much room for improvement in creative audio generation that primarily involves music and songs. Recent open lyrics-to-song models, such as, DiffRhythm, ACE-Step, and LeVo, have set an acceptable standard in automatic song generation for recreational use. However, these models lack fine-grained word-level controllability often desired by musicians in their workflows. To the best of our knowledge, our flow-matching-based JAM is the first effort toward endowing word-level timing and duration control in song generation, allowing fine-grained vocal control. To enhance the quality of generated songs to better align with human preferences, we implement aesthetic alignment through Direct Preference Optimization, which iteratively refines the model using a synthetic dataset, eliminating the need or manual data annotations. Furthermore, we aim to standardize the evaluation of such lyrics-to-song models through our public evaluation dataset JAME. We show that JAM outperforms the existing models in terms of the music-specific attributes.

cs.SD↗

arXiv2025

DOVA-PATBM: An Intelligent, Adaptive, and Scalable Framework for Optimizing Large-Scale EV Charging Infrastructure

The accelerating uptake of battery-electric vehicles demands infrastructure planning tools that are both data-rich and geographically scalable. Whereas most prior studies optimise charging locations for single cities, state-wide and national networks must reconcile the conflicting requirements of dense metropolitan cores, car-dependent exurbs, and power-constrained rural corridors. We present DOVA-PATBM (Deployment Optimisation with Voronoi-oriented, Adaptive, POI-Aware Temporal Behaviour Model), a geo-computational framework that unifies these contexts in a single pipeline. The method rasterises heterogeneous data (roads, population, night lights, POIs, and feeder lines) onto a hierarchical H3 grid, infers intersection importance with a zone-normalised graph neural network centrality model, and overlays a Voronoi tessellation that guarantees at least one five-port DC fast charger within every 30 km radius. Hourly arrival profiles, learned from loop-detector and floating-car traces, feed a finite M/M/c queue to size ports under feeder-capacity and outage-risk constraints. A greedy maximal-coverage heuristic with income-weighted penalties then selects the minimum number of sites that satisfy coverage and equity targets. Applied to the State of Georgia, USA, DOVA-PATBM (i) increases 30 km tile coverage by 12 percentage points, (ii) halves the mean distance that low-income residents travel to the nearest charger, and (iii) meets sub-transmission headroom everywhere -- all while remaining computationally tractable for national-scale roll-outs. These results demonstrate that a tightly integrated, GNN-driven, multi-resolution approach can bridge the gap between academic optimisation and deployable infrastructure policy.

cs.LG↗

arXiv2025

Lessons from Training Grounded LLMs with Verifiable Rewards

Generating grounded and trustworthy responses remains a key challenge for large language models (LLMs). While retrieval-augmented generation (RAG) with citation-based grounding holds promise, instruction-tuned models frequently fail even in straightforward scenarios: missing explicitly stated answers, citing incorrectly, or refusing when evidence is available. In this work, we explore how reinforcement learning (RL) and internal reasoning can enhance grounding in LLMs. We use the GRPO (Group Relative Policy Optimization) method to train models using verifiable outcome-based rewards targeting answer correctness, citation sufficiency, and refusal quality, without requiring gold reasoning traces or expensive annotations. Through comprehensive experiments across ASQA, QAMPARI, ELI5, and ExpertQA we show that reasoning-augmented models significantly outperform instruction-only variants, especially in handling unanswerable queries and generating well-cited responses. A two-stage training setup, first optimizing answer and citation behavior and then refusal, further improves grounding by stabilizing the learning signal. Additionally, we revisit instruction tuning via GPT-4 distillation and find that combining it with GRPO enhances performance on long-form, generative QA tasks. Overall, our findings highlight the value of reasoning, stage-wise optimization, and outcome-driven RL for building more verifiable and reliable LLMs.

cs.CL↗

arXiv2025

Machine Learning for Proactive Groundwater Management: Early Warning and Resource Allocation

Groundwater supports ecosystems, agriculture, and drinking water supplies worldwide, yet effective monitoring remains challenging due to sparse data, computational constraints, and delayed outputs from traditional approaches. We develop a machine learning pipeline that predicts groundwater level categories using climate data, hydro-meteorological records, and physiographic attributes processed through AutoGluon's automated ensemble framework. Our approach integrates geospatial preprocessing, domain-driven feature engineering, and automated model selection to overcome conventional monitoring limitations. Applied to a large-scale French dataset (n $>$ 3,440,000 observations from 1,500+ wells), the model achieves weighted F\_1 scores of 0.927 on validation data and 0.67 on temporally distinct test data. Scenario-based evaluations demonstrate practical utility for early warning systems and water allocation decisions under changing climate conditions. The open-source implementation provides a scalable framework for integrating machine learning into national groundwater monitoring networks, enabling more responsive and data-driven water management strategies.

eess.SP↗

arXiv2025

AttentionDrag: Exploiting Latent Correlation Knowledge in Pre-trained Diffusion Models for Image Editing

Traditional point-based image editing methods rely on iterative latent optimization or geometric transformations, which are either inefficient in their processing or fail to capture the semantic relationships within the image. These methods often overlook the powerful yet underutilized image editing capabilities inherent in pre-trained diffusion models. In this work, we propose a novel one-step point-based image editing method, named AttentionDrag, which leverages the inherent latent knowledge and feature correlations within pre-trained diffusion models for image editing tasks. This framework enables semantic consistency and high-quality manipulation without the need for extensive re-optimization or retraining. Specifically, we reutilize the latent correlations knowledge learned by the self-attention mechanism in the U-Net module during the DDIM inversion process to automatically identify and adjust relevant image regions, ensuring semantic validity and consistency. Additionally, AttentionDrag adaptively generates masks to guide the editing process, enabling precise and context-aware modifications with friendly interaction. Our results demonstrate a performance that surpasses most state-of-the-art methods with significantly faster speeds, showing a more efficient and semantically coherent solution for point-based image editing tasks.

cs.CV↗

arXiv2025

Latent Thought Models with Variational Bayes Inference-Time Computation

We propose a novel class of language models, Latent Thought Models (LTMs), which incorporate explicit latent thought vectors that follow an explicit prior model in latent space. These latent thought vectors guide the autoregressive generation of ground tokens through a Transformer decoder. Training employs a dual-rate optimization process within the classical variational Bayes framework: fast learning of local variational parameters for the posterior distribution of latent vectors (inference-time computation), and slow learning of global decoder parameters. Empirical studies reveal that LTMs possess additional scaling dimensions beyond traditional Large Language Models (LLMs), such as the number of iterations in inference-time computation and number of latent thought vectors. Higher sample efficiency can be achieved by increasing training compute per token, with further gains possible by trading model size for more inference steps. Designed based on these scaling properties, LTMs demonstrate superior sample and parameter efficiency compared to autoregressive models and discrete diffusion models. They significantly outperform these counterparts in validation perplexity and zero-shot language modeling tasks. Additionally, LTMs exhibit emergent few-shot in-context reasoning capabilities that scale with model size, and achieve competitive performance in conditional and unconditional text generation.

cs.CL↗

arXiv2025

Error Typing for Smarter Rewards: Improving Process Reward Models with Error-Aware Hierarchical Supervision

Large Language Models (LLMs) are prone to hallucination, especially during multi-hop and reasoning-intensive tasks such as mathematical problem solving. While Outcome Reward Models verify only final answers, Process Reward Models (PRMs) score each intermediate step to steer generation toward coherent solutions. We introduce PathFinder-PRM, a novel hierarchical, error-aware discriminative PRM that first classifies math and consistency errors at each step, then combines these fine-grained signals to estimate step correctness. To train PathFinder-PRM, we construct a 400K-sample dataset by enriching the human-annotated PRM800K corpus and RLHFlow Mistral traces with three-dimensional step-level labels. On PRMBench, PathFinder-PRM achieves a new state-of-the-art PRMScore of 67.7, outperforming the prior best (65.5) while using 3 times less data. When applied to reward guided greedy search, our model yields prm@8 48.3, a +1.5 point gain over the strongest baseline. These results demonstrate that decoupled error detection and reward estimation not only boost fine-grained error detection but also substantially improve end-to-end, reward-guided mathematical reasoning with greater data efficiency.

cs.CL↗

arXiv2025

A 2D Semantic-Aware Position Encoding for Vision Transformers

Vision transformers have demonstrated significant advantages in computer vision tasks due to their ability to capture long-range dependencies and contextual relationships through self-attention. However, existing position encoding techniques, which are largely borrowed from natural language processing, fail to effectively capture semantic-aware positional relationships between image patches. Traditional approaches like absolute position encoding and relative position encoding primarily focus on 1D linear position relationship, often neglecting the semantic similarity between distant yet contextually related patches. These limitations hinder model generalization, translation equivariance, and the ability to effectively handle repetitive or structured patterns in images. In this paper, we propose 2-Dimensional Semantic-Aware Position Encoding ($\text{SaPE}^2$), a novel position encoding method with semantic awareness that dynamically adapts position representations by leveraging local content instead of fixed linear position relationship or spatial coordinates. Our method enhances the model's ability to generalize across varying image resolutions and scales, improves translation equivariance, and better aggregates features for visually similar but spatially distant patches. By integrating $\text{SaPE}^2$ into vision transformers, we bridge the gap between position encoding and perceptual similarity, thereby improving performance on computer vision tasks.

cs.CV↗

arXiv2025

NORA: A Small Open-Sourced Generalist Vision Language Action Model for Embodied Tasks

Existing Visual-Language-Action (VLA) models have shown promising performance in zero-shot scenarios, demonstrating impressive task execution and reasoning capabilities. However, a significant challenge arises from the limitations of visual encoding, which can result in failures during tasks such as object grasping. Moreover, these models typically suffer from high computational overhead due to their large sizes, often exceeding 7B parameters. While these models excel in reasoning and task planning, the substantial computational overhead they incur makes them impractical for real-time robotic environments, where speed and efficiency are paramount. To address the limitations of existing VLA models, we propose NORA, a 3B-parameter model designed to reduce computational overhead while maintaining strong task performance. NORA adopts the Qwen-2.5-VL-3B multimodal model as its backbone, leveraging its superior visual-semantic understanding to enhance visual reasoning and action grounding. Additionally, our \model{} is trained on 970k real-world robot demonstrations and equipped with the FAST+ tokenizer for efficient action sequence generation. Experimental results demonstrate that NORA outperforms existing large-scale VLA models, achieving better task performance with significantly reduced computational overhead, making it a more practical solution for real-time robotic autonomy.

cs.RO↗

arXiv2025

Time-reversal symmetry breaking in microscopic single-crystal Sr$_2$RuO$_4$ devices

Time-reversal symmetry breaking superconductivity is a quintessential unconventional quantum state. In Josephson junctions, time-reversal symmetry breaking manifests itself in the supercurrent interference pattern as the invariance of the critical current under the reversal of both transport and magnetic field directions, i.e., $I_\text{c+}(H) = I_\text{c-}(-H)$. So far, such systems have been realized in devices where superconductivity is injected into a deliberately constructed weak link medium, usually carefully tuned by external magnetic fields and electrostatic gating. In this work, we report time-reversal symmetry breaking in spontaneously emerging Josephson junctions without intentionally constructed weak links. This is realized in ultra-pure single-crystal microstructures of Sr$_2$RuO$_4$, an unconventional superconductor with a multi-component order parameter. Here, the Josephson effect emerges intrinsically at the superconducting domain wall, where the degenerate states partially overlap. In addition to violating $I_\text{c+}(H) = I_\text{c-}(-H)$, we find a rich variety of exotic transport phenomena, including a supercurrent diode effect present in the entire interference pattern, two-channel critical current oscillations with a period that deviates from $Φ_0$, fractional Shapiro steps, and current-switchable bistable states with highly asymmetric critical currents. Our findings provide direct evidence of TRSB in unstrained Sr$_2$RuO$_4$ and reveal the potential of domain wall Josephson junctions, which can emerge in any superconductor where the pairing symmetry is described by a multi-component order parameter.

cond-mat.supr-con↗

arXiv2025

Exploring the origin of multi-periodic pulsations during a white-light flare

We explored the quasi-periodic pulsations (QPPs) at multiple periods during an X4.0 flare on 2024 May 10 (SOL2024-05-10T06:27), which occurred in the complex active region of NOAA 13664. The flare radiation reveals five prominent periods in multiple wavelengths. A 8-min QPP is simultaneously detected in wavelengths of HXR, radio, UV/EUV, Lya, and white light, which may be associated with nonthermal electrons periodically accelerated by intermittent magnetic reconnection that is modulated by the slow wave. A quasi-period at 14 minutes is observed in the SXR and high-temperature EUV wavebands, and it may be caused by repeatedly heated plasmas in hot flare loops. A quasiperiod at about 18 minutes is only observed by STIX, with reconstructed SXR images suggesting that the 18-min period pulsations should be considered as different flares. Meanwhile, a 3-min QPP is simultaneously detected in wavelengths of HXR, radio, and UV/ EUV, which is directly modulated by the slow magnetoacoustic wave leaking from sunspot umbrae. At last, a 2-min QPP is simultaneously detected in HXR and radio emissions during the pre-flare phase, which is possibly generated by a quasi-periodic regime of magnetic reconnection that is triggered by the kink wave.

astro-ph.SR↗

arXiv2025

TangoFlux: Super Fast and Faithful Text to Audio Generation with Flow Matching and Clap-Ranked Preference Optimization

We introduce TangoFlux, an efficient Text-to-Audio (TTA) generative model with 515M parameters, capable of generating up to 30 seconds of 44.1kHz audio in just 3.7 seconds on a single A40 GPU. A key challenge in aligning TTA models lies in the difficulty of creating preference pairs, as TTA lacks structured mechanisms like verifiable rewards or gold-standard answers available for Large Language Models (LLMs). To address this, we propose CLAP-Ranked Preference Optimization (CRPO), a novel framework that iteratively generates and optimizes preference data to enhance TTA alignment. We demonstrate that the audio preference dataset generated using CRPO outperforms existing alternatives. With this framework, TangoFlux achieves state-of-the-art performance across both objective and subjective benchmarks. We open source all code and models to support further research in TTA generation.

cs.SD↗

arXiv2025

PromptDistill: Query-based Selective Token Retention in Intermediate Layers for Efficient Large Language Model Inference

As large language models (LLMs) tackle increasingly complex tasks and longer documents, their computational and memory costs during inference become a major bottleneck. To address this, we propose PromptDistill, a novel, training-free method that improves inference efficiency while preserving generation quality. PromptDistill identifies and retains the most informative tokens by leveraging attention interactions in early layers, preserving their hidden states while reducing the computational burden in later layers. This allows the model to focus on essential contextual information without fully processing all tokens. Unlike previous methods such as H2O and SnapKV, which perform compression only after processing the entire input, or GemFilter, which selects a fixed portion of the initial prompt without considering contextual dependencies, PromptDistill dynamically allocates computational resources to the most relevant tokens while maintaining a global awareness of the input. Experiments using our method and baseline approaches with base models such as LLaMA 3.1 8B Instruct, Phi 3.5 Mini Instruct, and Qwen2 7B Instruct on benchmarks including LongBench, InfBench, and Needle in a Haystack demonstrate that PromptDistill significantly improves efficiency while having minimal impact on output quality compared to the original models. With a single-stage selection strategy, PromptDistill effectively balances performance and efficiency, outperforming prior methods like GemFilter, H2O, and SnapKV due to its superior ability to retain essential information. Specifically, compared to GemFilter, PromptDistill achieves an overall $1\%$ to $5\%$ performance improvement while also offering better time efficiency. Additionally, we explore multi-stage selection, which further improves efficiency while maintaining strong generation performance.

cs.CL↗

arXiv2025

Quantum transport in SnTe nanowire devices

We report on a variety of quantum transport experiments in SnTe nanowire devices. Research on these particular nanowire devices is relevant because of their topological properties and their potential to distinguish surface states owing to their high surface-to-volume ratio that suppresses the bulk contribution to the conductance. We observe a low-resistance and a high-resistance regime. The highly resistive devices display semiconducting and quantum dot behavior caused by microscopic differences in the fabrication, while devices with low resistance show partial superconductivity when in a hybrid superconductor-nanowire configuration or Fabry-Pérot oscillations. The latter suggests quantum interference in a ballistic transport channel, attributed to the 2D surface states in SnTe. The wide variety of quantum transport phenomena demonstrate SnTe nanowires as a promising platform for diverse follow-up experiments and novel device architectures, including the exploration of topological superconductivity and the development of low-energy spintronic devices.

cond-mat.mes-hall↗

arXiv2025

Minifilament Eruptions as the Last Straw to Break the Equilibrium of a Giant Solar Filament

Filament eruptions are magnetically driven violent explosions commonly observed on the Sun and late-type stars, sometimes leading to monster coronal mass ejections that directly affect the nearby planets' environments. More than a century of research on solar filaments suggests that the slow evolution of photospheric magnetic fields plays a decisive role in initiating filament eruptions, but the underlying mechanism remains unclear. Using high-resolution observations from the \textit{Chinese H$α$ Solar Explorer}, the \textit{Solar Upper Transition Region Imager}, and the \textit{Solar Dynamics Observatory}, we present direct evidence that a giant solar filament eruption is triggered by a series of minifilament eruptions occurring beneath it. These minifilaments, which are homologous to the giant filament but on a smaller tempo-spatial scale, sequently form and erupt due to extremely weak mutual flux disappearance of opposite-polarity photospheric magnetic fields. Through multi-fold magnetic interactions, these erupting minifilaments act as the last straw to break the force balance of the overlying giant filament and initiate its ultimate eruption. The results unveil a possible novel pathway for small-scale magnetic activities near the stellar surface to initiate spectacular filament eruptions, and provide new insight into the magnetic coupling of filament eruptions across different tempo-spatial scales.

astro-ph.SR↗

arXiv2025

From surface Fermi arcs to Fermi loops in the Dirac semimetal Cd3As2

Arc-like topological surface states, i.e., surface Fermi arcs, have long been recognized as the hallmark of Dirac semimetals. However, recent theories suggest that the surface Fermi arcs could evolve into closed Fermi loops, akin to surface states in topological insulators, while preserving the bulk Dirac semimetal phase. Here we experimentally reveal the evolution of Fermi arcs to Fermi loops in the surface-modified Dirac semimetal Cd3As2 nanoplate through gate voltage-dependent spin transport and quantum oscillation measurements. Surface modification, achieved by heavy metal atom deposition and water molecule adsorption, leads to an increase in the current-induced spin polarization at higher gate voltages, contrasting with the decrease observed in the pristine nanoplate. We also observe surface Shubnikov-de Haas oscillations with frequencies that scale linearly with gate voltage, aligning with a Fermi loop scenario. These findings indicate a transition from Fermi arcs to a closed Fermi loop in the surface-modified Cd3As2 nanoplate, consistent with the theoretically predicted fragile topological nature of Cd3As2. Our research offers profound insights into the transitions among these subtle topological states in Dirac semimetals, paving the way for manipulating topological surface states for high-performance spintronic devices.

cond-mat.mes-hall↗

arXiv2025

Goku: Flow Based Video Generative Foundation Models

This paper introduces Goku, a state-of-the-art family of joint image-and-video generation models leveraging rectified flow Transformers to achieve industry-leading performance. We detail the foundational elements enabling high-quality visual generation, including the data curation pipeline, model architecture design, flow formulation, and advanced infrastructure for efficient and robust large-scale training. The Goku models demonstrate superior performance in both qualitative and quantitative evaluations, setting new benchmarks across major tasks. Specifically, Goku achieves 0.76 on GenEval and 83.65 on DPG-Bench for text-to-image generation, and 84.85 on VBench for text-to-video tasks. We believe that this work provides valuable insights and practical advancements for the research community in developing joint image-and-video generation models.

cs.CV↗

arXiv2024

The magnetic origin of the mystery of rare H$α$ Moreton waves

Over the past three decades, a lot of coronal fast-mode waves were detected by space missions, but their counterparts in the chromosphere, called the Moreton waves, were rarely captured. How this happens remains a mystery. Here, to shed light on this problem, we investigate the photospheric vector magnetograms of the Moreton wave events associated with M- and X-class solar flares in 2010--2023. The H$α$ data are taken with the Global Oscillation Network Group (GONG) and the Chinese H$α$ Solar Explorer (CHASE). Our statistical results show that more than 80\% of the events occur at the edge of active regions and propagate non-radially due to asymmetric magnetic fields above the flares. According to the reconstructed magnetic field and atmospheric model, Moreton waves propagate in the direction along which the horizontal fast-mode wave speed drops the fastest. The result supports that the inclined magnetic configuration of the eruption is crucial to generate Moreton waves, even for X-class flares. It may explain the low occurrence rate of Moreton waves and why some X-class flares accompanied with coronal mass ejections (CMEs) do not generate Moreton waves.

astro-ph.SR↗