arXiv subjects

Chen Zhang

Publications and source records attributed to Chen Zhang.

At least 271 records · Page 15Linked to original sources

arXiv2025

Towards Flow-Matching-based TTS without Classifier-Free Guidance

Flow matching has demonstrated strong generative capabilities and has become a core component in modern Text-to-Speech (TTS) systems. To ensure high-quality speech synthesis, Classifier-Free Guidance (CFG) is widely used during the inference of flow-matching-based TTS models. However, CFG incurs substantial computational cost as it requires two forward passes, which hinders its applicability in real-time scenarios. In this paper, we explore removing CFG from flow-matching-based TTS models to improve inference efficiency, while maintaining performance. Specifically, we reformulated the flow matching training target to directly approximate the CFG optimization trajectory. This training method eliminates the need for unconditional model evaluation and guided tuning during inference, effectively cutting the computational overhead in half. Furthermore, It can be seamlessly integrated with existing optimized sampling strategies. We validate our approach using the F5-TTS model on the LibriTTS dataset. Experimental results show that our method achieves a 9$\times$ inference speed-up compared to the baseline F5-TTS, while preserving comparable speech quality. We will release the code and models to support reproducibility and foster further research in this area.

eess.AS↗

arXiv2025

CICADA: Cross-Domain Interpretable Coding for Anomaly Detection and Adaptation in Multivariate Time Series

Unsupervised Time series anomaly detection plays a crucial role in applications across industries. However, existing methods face significant challenges due to data distributional shifts across different domains, which are exacerbated by the non-stationarity of time series over time. Existing models fail to generalize under multiple heterogeneous source domains and emerging unseen new target domains. To fill the research gap, we introduce CICADA (Cross-domain Interpretable Coding for Anomaly Detection and Adaptation), with four key innovations: (1) a mixture of experts (MOE) framework that captures domain-agnostic anomaly features with high flexibility and interpretability; (2) a novel selective meta-learning mechanism to prevent negative transfer between dissimilar domains, (3) an adaptive expansion algorithm for emerging heterogeneous domain expansion, and (4) a hierarchical attention structure that quantifies expert contributions during fusion to enhance interpretability further.Extensive experiments on synthetic and real-world industrial datasets demonstrate that CICADA outperforms state-of-the-art methods in both cross-domain detection performance and interpretability.

cs.LG↗

arXiv2025

From Human Memory to AI Memory: A Survey on Memory Mechanisms in the Era of LLMs

Memory is the process of encoding, storing, and retrieving information, allowing humans to retain experiences, knowledge, skills, and facts over time, and serving as the foundation for growth and effective interaction with the world. It plays a crucial role in shaping our identity, making decisions, learning from past experiences, building relationships, and adapting to changes. In the era of large language models (LLMs), memory refers to the ability of an AI system to retain, recall, and use information from past interactions to improve future responses and interactions. Although previous research and reviews have provided detailed descriptions of memory mechanisms, there is still a lack of a systematic review that summarizes and analyzes the relationship between the memory of LLM-driven AI systems and human memory, as well as how we can be inspired by human memory to construct more powerful memory systems. To achieve this, in this paper, we propose a comprehensive survey on the memory of LLM-driven AI systems. In particular, we first conduct a detailed analysis of the categories of human memory and relate them to the memory of AI systems. Second, we systematically organize existing memory-related work and propose a categorization method based on three dimensions (object, form, and time) and eight quadrants. Finally, we illustrate some open problems regarding the memory of current AI systems and outline possible future directions for memory in the era of large language models.

cs.IR↗

arXiv2025

HistLLM: A Unified Framework for LLM-Based Multimodal Recommendation with User History Encoding and Compression

While large language models (LLMs) have proven effective in leveraging textual data for recommendations, their application to multimodal recommendation tasks remains relatively underexplored. Although LLMs can process multimodal information through projection functions that map visual features into their semantic space, recommendation tasks often require representing users' history interactions through lengthy prompts combining text and visual elements, which not only hampers training and inference efficiency but also makes it difficult for the model to accurately capture user preferences from complex and extended prompts, leading to reduced recommendation performance. To address this challenge, we introduce HistLLM, an innovative multimodal recommendation framework that integrates textual and visual features through a User History Encoding Module (UHEM), compressing multimodal user history interactions into a single token representation, effectively facilitating LLMs in processing user preferences. Extensive experiments demonstrate the effectiveness and efficiency of our proposed mechanism.

cs.IR↗

arXiv2025

Partition laser assembling technique

The advancement of micro/nanofabrication techniques with high throughput, efficiency, and flexibility is critical for fields like integrated photonics, biosensing, and medical diagnostics. This study presents Partition Laser Assembling (PLA), a novel laser technique for fabricating complex micro/nanostructures akin to puzzle pieces. By dividing the target patterns described by scalable vector graphics into partitions, any structures in each partition can be fabricated via structured lights as "light stamp" through spatial light modulation. Unlike traditional direct laser writing, PLA eliminates reliance on mechanical components, avoiding step-like artifacts and ensuring smoother fabrication of complex micro/nanostructures. By seamlessly assembling basic shapes, PLA achieves intricate structures like micro artworks and metalenses with unmatched precision and resolution. Leveraging two-photon fabrication, PLA guarantees high resolution and structural integrity, positioning it as a transformative tool for nanoscale 3D printing. With applications spanning research and industry, PLA paves the way for advanced optical devices, micro/nanofabrications, and next-gen manufacturing technologies.

physics.optics↗

arXiv2025

Rotational Properties of Inverted Hybrid Stars

We study the rotational properties of inverted hybrid stars (also termed cross stars), which have been recently proposed as a possible new class of compact stars characterized by an outer layer of quark matter and a core of hadrons, in an inverted structure compared to traditional hybrid stars. We analyze distinct models representing varying depths of quark-hadron phase transitions. Our findings reveal that, while cross stars rotating at their Kepler frequencies typically exhibit a significantly higher mass and larger circumferential radius as anticipated, interestingly, there is a significant increase in potential twin configurations in the case of rapid rotations. We further study sequences of constant baryonic mass, representing potential paths of rotational evolution. Our results indicate that not all stars in these sequences are viable due to the onset of phase transitions during spin-down, leading to possible mini-collapses. We also investigate the phenomenon of ``back-bending" during spin-down sequences, which is manifested in a rather different shape for cross stars due to their inverted structure and the large density discontinuity caused by the strong phase transition, in contrast to traditional hybrid stars. Our research enriches existing studies by introducing the significant aspect of rotation, unveiling intr

astro-ph.HE↗

arXiv2025

DOMAC: Differentiable Optimization for High-Speed Multipliers and Multiply-Accumulators

Multipliers and multiply-accumulators (MACs) are fundamental building blocks for compute-intensive applications such as artificial intelligence. With the diminishing returns of Moore's Law, optimizing multiplier performance now necessitates process-aware architectural innovations rather than relying solely on technology scaling. In this paper, we introduce DOMAC, a novel approach that employs differentiable optimization for designing multipliers and MACs at specific technology nodes. DOMAC establishes an analogy between optimizing multi-staged parallel compressor trees and training deep neural networks. Building on this insight, DOMAC reformulates the discrete optimization challenge into a continuous problem by incorporating differentiable timing and area objectives. This formulation enables us to utilize existing deep learning toolkit for highly efficient implementation of the differentiable solver. Experimental results demonstrate that DOMAC achieves significant enhancements in both performance and area efficiency compared to state-of-the-art baselines and commercial IPs in multiplier and MAC designs.

cs.AR↗

arXiv2025

DiffuSE: Cross-Layer Design Space Exploration of DNN Accelerator via Diffusion-Driven Optimization

The proliferation of deep learning accelerators calls for efficient and cost-effective hardware design solutions, where parameterized modular hardware generator and electronic design automation (EDA) tools play crucial roles in improving productivity and final Quality-of-Results (QoR). To strike a good balance across multiple QoR of interest (e.g., performance, power, and area), the designers need to navigate a vast design space, encompassing tunable parameters for both hardware generator and EDA synthesis tools. However, the significant time for EDA tool invocations and complex interplay among numerous design parameters make this task extremely challenging, even for experienced designers. To address these challenges, we introduce DiffuSE, a diffusion-driven design space exploration framework for cross-layer optimization of DNN accelerators. DiffuSE leverages conditional diffusion models to capture the inverse, one-to-many mapping from QoR objectives to parameter combinations, allowing for targeted exploration within promising regions of the design space. By carefully selecting the conditioning QoR values, the framework facilitates an effective trade-off among multiple QoR metrics in a sample-efficient manner. Experimental results under 7nm technology demonstrate the superiority of the proposed framework compared to previous arts.

cs.AR↗

arXiv2025

MegaTTS 3: Sparse Alignment Enhanced Latent Diffusion Transformer for Zero-Shot Speech Synthesis

While recent zero-shot text-to-speech (TTS) models have significantly improved speech quality and expressiveness, mainstream systems still suffer from issues related to speech-text alignment modeling: 1) models without explicit speech-text alignment modeling exhibit less robustness, especially for hard sentences in practical applications; 2) predefined alignment-based models suffer from naturalness constraints of forced alignments. This paper introduces \textit{MegaTTS 3}, a TTS system featuring an innovative sparse alignment algorithm that guides the latent diffusion transformer (DiT). Specifically, we provide sparse alignment boundaries to MegaTTS 3 to reduce the difficulty of alignment without limiting the search space, thereby achieving high naturalness. Moreover, we employ a multi-condition classifier-free guidance strategy for accent intensity adjustment and adopt the piecewise rectified flow technique to accelerate the generation process. Experiments demonstrate that MegaTTS 3 achieves state-of-the-art zero-shot TTS speech quality and supports highly flexible control over accent intensity. Notably, our system can generate high-quality one-minute speech with only 8 sampling steps. Audio samples are available at https://sditdemo.github.io/sditdemo/.

eess.AS↗

arXiv2025

WindowKV: Task-Adaptive Group-Wise KV Cache Window Selection for Efficient LLM Inference

With the advancements in long-context inference capabilities of large language models (LLMs), the KV cache has become one of the foundational components. However, its substantial GPU memory consumption makes KV cache compression a key technique for enabling efficient LLM inference in industrial scenarios. While recent studies have focused on optimizing the memory occupied by the KV cache, they overlook two critical factors: preserving semantic coherence and considering task-specific characteristic during compression. To address these limitations, we propose a novel task-adaptive KV cache window selection method, WindowKV. WindowKV dynamically selects local semantic windows consisting of consecutive tokens, according to task-specific characteristics, ensuring the retained KV cache captures continuous, essential context. Additionally, we introduce an intra-group layer KV cache indices sharing strategy to reduce computational overhead, achieving a balance between performance and efficiency. We rigorously evaluate WindowKV on the LongBench benchmark, and the results demonstrate that it maintains a performance comparable to full KV cache retention while using only 12% of the original KV cache, significantly reducing memory requirements. Furthermore, our method also achieves state-of-the-art results in the Needle-in-a-Haystack evaluation, highlighting its effectiveness and robustness.

cs.CL↗

arXiv2025

Ultrafast photoinduced phase transition in the antiferromagnetic Dirac semimetal EuAgAs

We report the observation of a light-induced subpicosecond phase transition in the antiferromagnetic Dirac semimetal EuAgAs, achieved through ultrafast optical excitation. Using ultrafast optical spectroscopy, we probe the nonequilibrium carrier dynamics, discovering distinct fluence-dependent responses in the antiferromagnetic and paramagnetic states, and revealing a possible magnetic order-driven transition between different topological states. Our results demonstrate that EuAgAs, with its highly tunable magnetic structure, possibly offers a unique platform for exploring topological phase transitions. These results underscore the potential of ultrashort optical pulses as powerful tools for the real-time control of topological phases, opening pathways for advances in spintronics, quantum computing, and energy-efficient information technologies.

cond-mat.str-el↗

arXiv2025

Jenga: Effective Memory Management for Serving LLM with Heterogeneity

Large language models (LLMs) are widely used but expensive to run, especially as inference workloads grow. To lower costs, maximizing the request batch size by managing GPU memory efficiently is crucial. While PagedAttention has recently been proposed to improve the efficiency of memory management, we find that the growing heterogeneity in the embeddings dimensions, attention, and access patterns of modern LLM architectures introduces new challenges for memory allocation. In this paper, we present Jenga, a novel memory allocation framework for heterogeneous embeddings in LLMs. Jenga tackles two key challenges: (1) minimizing memory fragmentation when managing embeddings of different sizes, and (2) enabling flexible caching and eviction policies tailored to the specific token-dependency patterns of various layers. Jenga employs a two-level memory allocator, leveraging the least common multiple (LCM) of embedding sizes to optimize memory usage and providing APIs to express layer-specific caching logic to enhance memory reuse. We implemente Jenga on vLLM, a state-of-the-art LLM inference engine, and evaluate it with diverse LLMs, datasets, and GPU configurations. Evaluations show that Jenga improves GPU memory utilization by up to 79.6%, and increases serving throughput by up to 4.92x (1.80x on average).

cs.DC↗

arXiv2025

Anisotropic Hybridization Dynamics in the Quasi-One-Dimensional Kondo Lattice CeCo$_2$Ga$_8$ Revealed by Ultrafast Optical Spectroscopy

We investigate the ultrafast dynamics of the quasi-one-dimensional Kondo lattice CeCo$_2$Ga$_8$ using optical pump-probe spectroscopy. Time-resolved pump-probe reflectivity measurements reveal a strong anisotropy in the photoinduced response, which is a direct consequence of the material's unique electronic structure. The temperature dependence of the relaxation dynamics provides evidence for the formation of two distinct hybridization gaps that appear at different temperatures in the heavy fermion state. A direct gap of 2$Δ_{dir}$ $\approx$ 50 meV that persists up to $T^†$ $\approx$ 90 K, well above the coherence temperature $T^*$ $\approx$ 20 K. We attribute this higher-temperature gap to the hybridization fluctuations. An indirect gap of 2$Δ_{ind}$ $\approx$ 10 meV opens closer to $T^*$, signifying the development of long-range coherence in the heavy fermion state. Furthermore, we find that the hybridization gap can be suppressed with increasing pump fluence, indicating a delicate interplay between photoexcitation and the coherent heavy fermion state. Our results provide insights into the interplay of Kondo physics and low dimensionality in CeCo$_2$Ga$_8$, and establish ultrafast optical spectroscopy as a sensitive probe of anisotropic hybridization in heavy fermion materials.

cond-mat.str-el↗

arXiv2025

Interplay of electron-phonon coupling, pseudogap, and superconductivity in CsCa$_2$Fe$_4$As$_4$F$_2$ studied using ultrafast optical spectroscopy

The quasiparticle relaxation dynamics of the iron-based superconductor CsCa$_2$Fe$_4$As$_4$F$_2$ ($T_c$ $\sim$ 29 K) were investigated using ultrafast optical spectroscopy. A pseudogap ($Δ_{PG}$ $\approx$ 3.3 meV) was observed to open below $T^{\ast}$ $\approx$ 60 K, prior to the emergence of a superconducting gap ($Δ$ $\approx$ 6.6 $\pm$ 0.4 meV). At high excitation fluence, a coherent A1g phonon mode at 5.49 THz was identified, exhibiting deviations from anharmonic behavior below $T_c$. The electron-phonon coupling constant for this mode was estimated to be $λ_{A_{1g}}$ $\approx$ 0.23. These results provide insights into the interplay between the electron-phonon interactions, pseudogap, and the superconducting pairing mechanism in CsCa$_2$Fe$_4$As$_4$F$_2$.

cond-mat.supr-con↗

arXiv2025

Bayesian inference of strangeon matter using the measurements of PSR J0437-4715 and GW190814

The observations of compact star inspirals from LIGO/Virgo combined with mass and radius measurements from NICER provide a valuable tool to study the highly uncertain equation of state (EOS) of dense matter at the densities characteristic of compact stars. In this work, we constrain the solid states of strange-cluster matter, called strangeon matter, as the putative basic units of the ground state of bulk strong matter using a Bayesian statistical method, incorporating the mass and radius measurements of PSR J0030+0451, PSR J0740+6620, and the recent data for the $1.4\ M_{\odot}$ pulsar PSR J0437-4715. We also include constraints from gravitational wave events GW170817 and GW190814. Under the prior assumption of a finite number of quarks in a strangeon, $N_{\rm q}$, our analysis reveals that current mass-radius measurements favor a larger $N_{\rm q}$. Specifically, the results support the scenario where a strangeon forms a stable bound state with $N_{\rm q}=18$, symmetric in color, flavor, and spin spaces, compared to the minimum $N_{\rm q}$ prior. The comparative analyses of the posterior EOS parameter spaces derived from three-parameter model and two-parameter model demonstrate a consistent prediction under identical observational constraints. In particular, our results indicate that the most probable values of the maximum mass are found to be $3.58^{+0.16}_{-0.12}\ M_{\odot}$ ($3.65^{+0.18}_{-0.16}\ M_{\odot}$) at $90\%$ confidence level for three-parameter (two-parameter) EOS considering the constraints of GW190814. The corresponding radii for $1.4\ M_{\odot}$ and $2.1\ M_{\odot}$ stars are $12.04^{+0.27}_{-0.31}~\rm km$ ($12.16^{+0.26}_{-0.31}~\rm km$) and $13.43^{+0.31}_{-0.32}~\rm km$ ($13.60^{+0.29}_{-0.34}~\rm km$), respectively. This result may impact interestingly on the research of multiquark states, which could improve our understanding of the nonperturbative strong force.

astro-ph.HE↗

arXiv2025

Programmable Electric Tweezers

The interaction mechanism between a single microscopic object like a cell, a particle, a molecule, or an atom and its interacting electromagnetic field is fundamental in single-object manipulation such as optical trap and magnetic trap. Function-on-demand, single-object manipulation relies on a high degree of freedom control of electromagnetic field at localized scales, which remains challenging. Here we propose a manipulation concept: programmable single-object manipulation, based on programming the electromagnetic field in a multi-bit electrode system. This concept is materialized on a Programmable Electric Tweezer (PET) with four individually addressed electrodes, marking a transition from function-fixed single-object manipulation to function-programmable single-object manipulation. By programming the localized electric field, our PET can provide various manipulation functions for achieving precise trapping, movement and rotation of multiscale single microscopic objects, including single proteins, nucleic acids, microparticles and bacteria. Implementing these functions, we are able not only to manipulate the object of interest on demand but also quantitatively measure the charge to mass ratio of a single microparticle via the Paul trap and the electrical properties of an individual bacterial cell by the rotation analysis. Finally, with superposed single-particle trapping and rotation, we demonstrate the spontaneous relaxation of DNA supercoiling and observe an unexpected pause phenomenon in the relaxation process, highlighting the versatility and the potential of PET in uncovering stochastic biophysical phenomena at the single-molecule level.

physics.bio-ph↗

arXiv2025

Aligning Language Models Using Follow-up Likelihood as Reward Signal

In natural human-to-human conversations, participants often receive feedback signals from one another based on their follow-up reactions. These reactions can include verbal responses, facial expressions, changes in emotional state, and other non-verbal cues. Similarly, in human-machine interactions, the machine can leverage the user's follow-up utterances as feedback signals to assess whether it has appropriately addressed the user's request. Therefore, we propose using the likelihood of follow-up utterances as rewards to differentiate preferred responses from less favored ones, without relying on human or commercial LLM-based preference annotations. Our proposed reward mechanism, ``Follow-up Likelihood as Reward" (FLR), matches the performance of strong reward models trained on large-scale human or GPT-4 annotated data on 8 pairwise-preference and 4 rating-based benchmarks. Building upon the FLR mechanism, we propose to automatically mine preference data from the online generations of a base policy model. The preference data are subsequently used to boost the helpfulness of the base model through direct alignment from preference (DAP) methods, such as direct preference optimization (DPO). Lastly, we demonstrate that fine-tuning the language model that provides follow-up likelihood with natural language feedback significantly enhances FLR's performance on reward modeling benchmarks and effectiveness in aligning the base policy model's helpfulness.

cs.CL↗

arXiv2025

LEIA discovery of the longest-lasting and most energetic stellar X-ray flare ever detected

The Lobster Eye Imager for Astronomy (LEIA) detected a new X-ray transient on 2022 November 7, identified as a superflare event occurring on a nearby K-type giant star HD 251108. The flux increase was also detected in follow-up observations at X-ray, UV and optical wavelengths. The flare lasted for about 40 days in soft X-ray observations, reaching a peak luminosity of ~1.1 * 10^34 erg/s in 0.5-4.0 keV, which is roughly 60 times the quiescent luminosity. Optical brightening was observed for only one night. The X-ray light curve is well described by a double fast rise and exponential decay model, attributed to the cooling process of a loop arcade structure formed subsequent to the initial large loop with a half-length of ~1.9 * 10^12 cm. Time-resolved X-ray spectra were fitted by a four-temperature apec model (with three components being the quiescent background), showing significant evolution of plasma temperature and emission measure over time. The estimated energy released in the LEIA band is ~3 * 10^39 erg, suggesting that this is likely the most energetic X-ray stellar flare with the longest duration detected to date.

astro-ph.HE↗