arXiv subjects

Bin Chen

Publications and source records attributed to Bin Chen.

At least 235 records · Page 13Linked to original sources

arXiv2025

MQAD: A Large-Scale Question Answering Dataset for Training Music Large Language Models

Question-answering (QA) is a natural approach for humans to understand a piece of music audio. However, for machines, accessing a large-scale dataset covering diverse aspects of music is crucial, yet challenging, due to the scarcity of publicly available music data of this type. This paper introduces MQAD, a music QA dataset built on the Million Song Dataset (MSD), encompassing a rich array of musical features, including beat, chord, key, structure, instrument, and genre -- across 270,000 tracks, featuring nearly 3 million diverse questions and captions. MQAD distinguishes itself by offering detailed time-varying musical information such as chords and sections, enabling exploration into the inherent structure of music within a song. To compile MQAD, our methodology leverages specialized Music Information Retrieval (MIR) models to extract higher-level musical features and Large Language Models (LLMs) to generate natural language QA pairs. Then, we leverage a multimodal LLM that integrates the LLaMA2 and Whisper architectures, along with novel subjective metrics to assess the performance of MQAD. In experiments, our model trained on MQAD demonstrates advancements over conventional music audio captioning approaches. The dataset and code are available at https://github.com/oyzh888/MQAD.

cs.SD↗

arXiv2025

Semi-analytical Study on the Polarized Images of Black Hole due to Frame Dragging

An initially retrograde accretion flow is transformed into a prograde configuration before plunging into the black hole, as a result of the frame-dragging effect induced by the black hole's rotation. The polarized image of a black hole shaped by such an accretion flow manifests three distinctive critical locations: the turning point of the flow's primary image, the polarization-flip location on the image plane, and the position of the primary image corresponding to the flow's actual turning point in spacetime. Due to the influences of gravitational lensing and gravitational Faraday rotation, these three positions generally do not coincide. In this work, we examine a thin equatorial accretion disk composed of initially retrograde, geodesically moving flows, and conduct a systematic investigation into the interrelations and discrepancies among these critical locations. We elucidate the spatial hierarchy among the three, and for an on-axis observer, we derive approximate analytic expressions characterizing their positions.

gr-qc↗

arXiv2025

Energy Conversion and Electron Acceleration and Transport in 3D Simulations of Solar Flares

Recent observations and simulations indicate that solar flares undergo extremely complex three-dimensional (3D) evolution, making 3D particle transport models essential for understanding electron acceleration and interpreting flare emissions. In this study, we investigate this problem by solving Parker's transport equation with 3D MHD simulations of solar flares. By examining energy conversion in the 3D system, we evaluate the roles of different acceleration mechanisms, including reconnection current sheet (CS), termination shock (TS), and supra-arcade downflows (SADs). We find that large-amplitude turbulent fluctuations are generated and sustained in the 3D system. The model results demonstrate that a significant number of electrons are accelerated to hundreds of keV and even a few MeV, forming power-law energy spectra. These energetic particles are widely distributed, with concentrations at the TS and in the flare looptop region, consistent with results derived from recent hard X-ray (HXR) and microwave (MW) observations. By selectively turning particle acceleration on or off in specific regions, we find that the CS and SADs effectively accelerate electrons to several hundred keV, while the TS enables further acceleration to MeV. However, no single mechanism can independently account for the significant number of energetic electrons observed. Instead, the mechanisms work synergistically to produce a large population of accelerated electrons. Our model provides spatially and temporally resolved electron distributions in the whole flare region and at the flare footpoints, enabling synthetic HXR and MW emission modeling for comparison with observations. These results offer important insights into electron acceleration and transport in 3D solar flare regions.

astro-ph.SR↗

arXiv2025

Generalized Decoupled Learning for Enhancing Open-Vocabulary Dense Perception

Dense visual perception tasks have been constrained by their reliance on predefined categories, limiting their applicability in real-world scenarios where visual concepts are unbounded. While Vision-Language Models (VLMs) like CLIP have shown promise in open-vocabulary tasks, their direct application to dense perception often leads to suboptimal performance due to limitations in local feature representation. In this work, we present our observation that CLIP's image tokens struggle to effectively aggregate information from spatially or semantically related regions, resulting in features that lack local discriminability and spatial consistency. To address this issue, we propose DeCLIP, a novel framework that enhances CLIP by decoupling the self-attention module to obtain ``content'' and ``context'' features respectively. \revise{The context features are enhanced by jointly distilling semantic correlations from Vision Foundation Models (VFMs) and object integrity cues from diffusion models, thereby enhancing spatial consistency. In parallel, the content features are aligned with image crop representations and constrained by region correlations from VFMs to improve local discriminability. Extensive experiments demonstrate that DeCLIP establishes a solid foundation for open-vocabulary dense perception, consistently achieving state-of-the-art performance across a broad spectrum of tasks, including 2D detection and segmentation, 3D instance segmentation, video instance segmentation, and 6D object pose estimation.} Code is available at https://github.com/xiaomoguhz/DeCLIP

cs.CV↗

arXiv2025

SeamlessFlow: A Trainer Agent Isolation RL Framework Achieving Bubble-Free Pipelines via Tag Scheduling

We introduce SeamlessFlow, a server based reinforcement learning (RL) framework that addresses two core challenges in industrial scale RL: (1) decoupling RL training from the complex execution flow of agents; (2) maximizing GPU utilization with minimal idle time while preserving the stability and scalability required for large-scale deployments. First, SeamlessFlow introduces a data plane that decouples the RL trainer from diverse, complex agent implementations while sustaining high throughput. A central trajectory manager maintains complete interaction histories and supports partial rollout, allowing rollout to pause for weight updates and resume seamlessly, keeping agents unaware of service interruptions. Second, we propose a tag driven scheduling paradigm that abstracts hardware into capability tagged resources, unifying colocated and disaggregated architectures. Based on this, SeamlessFlow introduces a spatiotemporal multiplexing pipeline that dynamically reassigns idle training nodes to rollout in a train rollout separated setup, eliminating pipeline bubbles and fully exploiting heterogeneous cluster resources. By combining these innovations, SeamlessFlow delivers both stability and high performance, making it well suited for multi agent, long horizon, and other complex RL tasks.

cs.LG↗

arXiv2025

GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models

We present GLM-4.5, an open-source Mixture-of-Experts (MoE) large language model with 355B total parameters and 32B activated parameters, featuring a hybrid reasoning method that supports both thinking and direct response modes. Through multi-stage training on 23T tokens and comprehensive post-training with expert model iteration and reinforcement learning, GLM-4.5 achieves strong performance across agentic, reasoning, and coding (ARC) tasks, scoring 70.1% on TAU-Bench, 91.0% on AIME 24, and 64.2% on SWE-bench Verified. With much fewer parameters than several competitors, GLM-4.5 ranks 3rd overall among all evaluated models and 2nd on agentic benchmarks. We release both GLM-4.5 (355B parameters) and a compact version, GLM-4.5-Air (106B parameters), to advance research in reasoning and agentic AI systems. Code, models, and more information are available at https://github.com/zai-org/GLM-4.5.

cs.CL↗

arXiv2025

Neighborhood-Preserving Voronoi Treemaps

Voronoi treemaps are used to depict nodes and their hierarchical relationships simultaneously. However, in addition to the hierarchical structure, data attributes, such as co-occurring features or similarities, frequently exist. Examples include geographical attributes like shared borders between countries or contextualized semantic information such as embedding vectors derived from large language models. In this work, we introduce a Voronoi treemap algorithm that leverages data similarity to generate neighborhood-preserving treemaps. First, we extend the treemap layout pipeline to consider similarity during data preprocessing. We then use a Kuhn-Munkres matching of similarities to centroidal Voronoi tessellation (CVT) cells to create initial Voronoi diagrams with equal cell sizes for each level. Greedy swapping is used to improve the neighborhoods of cells to match the data's similarity further. During optimization, cell areas are iteratively adjusted to their respective sizes while preserving the existing neighborhoods. We demonstrate the practicality of our approach through multiple real-world examples drawn from infographics and linguistics. To quantitatively assess the resulting treemaps, we employ treemap metrics and measure neighborhood preservation.

cs.GR↗

arXiv2025

Carrollian superstring in the flipped vacuum

In this work, we study the spectrum of the Carrollian superstring in the flipped vacuum (the highest-weight representation) in detail. The target spacetime of the Carrollian string has a generalized Carrollian symmetry, and it is composed of both Poincaré directions and Carrollian directions. We explicitly show that two homogeneous Carrollian superstring theories, one with compactified Poincaré direction and the other with compactified Carrollian direction, share the same BMS$_3$ symmetry, and their zero modes can be identified under the usual T-duality transformation. Moreover, we investigate the spectrum of a general Carrollian superstring in the flipped vacuum. As the string can still have nonvanishing winding along the spatial direction in the infinite radius limit, the spectrum of the Carrollian strings in the flat background is no longer truncated. We furthermore construct the vertex operators of gravitons and discuss their scattering amplitudes. We find that the form of 3-point amplitudes differs from that of the usual tensile superstrings only by a simple function, which reflects the ultra-local nature of Carrollian physics.

hep-th↗

arXiv2025

Root-$T\bar{T}$ Flows Unify 4D Duality-Invariant Electrodynamics and 2D Integrable Sigma Models

We present a unified framework that connects four-dimensional duality-invariant nonlinear electrodynamics and two-dimensional integrable sigma models via the Courant-Hilbert and new auxiliary field formulations, both governed by a common generating function and a generating potential, respectively. Introducing two commuting deformation parameters, $λ$ (irrelevant) and $γ$ (marginal), we identify a universal class of $γ$-flows, including the root-$T\bar{T}$ deformation and its rescaled variants. Our approach generalizes conventional single-coupling structures via novel field transformations that extend to a two-parameter space ($λ$,$γ$) while preserving the root-$T\bar{T}$ flow condition for all $γ$-coupled theories. We construct several integrable models, including generalized Born-Infeld, logarithmic, q-deformed, and a new closed-form theory applicable to both electrodynamics and integrable systems. This unified framework, based on the unique form of the root-$T\bar{T}$ flow, systematically spans duality-invariant nonlinear electrodynamics in 4D and their exact 2D integrable counterparts.

hep-th↗

arXiv2025

HLFormer: Enhancing Partially Relevant Video Retrieval with Hyperbolic Learning

Partially Relevant Video Retrieval (PRVR) addresses the critical challenge of matching untrimmed videos with text queries describing only partial content. Existing methods suffer from geometric distortion in Euclidean space that sometimes misrepresents the intrinsic hierarchical structure of videos and overlooks certain hierarchical semantics, ultimately leading to suboptimal temporal modeling. To address this issue, we propose the first hyperbolic modeling framework for PRVR, namely HLFormer, which leverages hyperbolic space learning to compensate for the suboptimal hierarchical modeling capabilities of Euclidean space. Specifically, HLFormer integrates the Lorentz Attention Block and Euclidean Attention Block to encode video embeddings in hybrid spaces, using the Mean-Guided Adaptive Interaction Module to dynamically fuse features. Additionally, we introduce a Partial Order Preservation Loss to enforce "text < video" hierarchy through Lorentzian cone constraints. This approach further enhances cross-modal matching by reinforcing partial relevance between video content and text queries. Extensive experiments show that HLFormer outperforms state-of-the-art methods. Code is released at https://github.com/lijun2005/ICCV25-HLFormer.

cs.CV↗

arXiv2025

EDPC: Accelerating Lossless Compression via Lightweight Probability Models and Decoupled Parallel Dataflow

The explosive growth of multi-source multimedia data has significantly increased the demands for transmission and storage, placing substantial pressure on bandwidth and storage infrastructures. While Autoregressive Compression Models (ACMs) have markedly improved compression efficiency through probabilistic prediction, current approaches remain constrained by two critical limitations: suboptimal compression ratios due to insufficient fine-grained feature extraction during probability modeling, and real-time processing bottlenecks caused by high resource consumption and low compression speeds. To address these challenges, we propose Efficient Dual-path Parallel Compression (EDPC), a hierarchically optimized compression framework that synergistically enhances modeling capability and execution efficiency via coordinated dual-path operations. At the modeling level, we introduce the Information Flow Refinement (IFR) metric grounded in mutual information theory, and design a Multi-path Byte Refinement Block (MBRB) to strengthen cross-byte dependency modeling via heterogeneous feature propagation. At the system level, we develop a Latent Transformation Engine (LTE) for compact high-dimensional feature representation and a Decoupled Pipeline Compression Architecture (DPCA) to eliminate encoding-decoding latency through pipelined parallelization. Experimental results demonstrate that EDPC achieves comprehensive improvements over state-of-the-art methods, including a 2.7x faster compression speed, and a 3.2% higher compression ratio. These advancements establish EDPC as an efficient solution for real-time processing of large-scale multimedia data in bandwidth-constrained scenarios. Our code is available at https://github.com/Magie0/EDPC.

cs.IT↗

arXiv2025

Reflected entropy in an evaporating black hole through non-isometric map

The black hole information paradox has been an important problem in quantum gravity. In the study of evaporating black hole, it has been proposed that the holographic map between the semi-classical effective description in bulk and the fundamental description in boundary cannot be isometric. In this work, we would like to study the reflected entropy in an evaporating black hole model through non-isometric holographic map. We assume that the evaporating is slowly enough that it makes sense to ascribe a slowly varying temperature to the Hawking radiation. We then introduce a two-sided black hole model to canonically purify the semi-classical state. The holographic map to the fundamental description is non-isometric and defined by a Haar random unitary matrix. We show that the entanglement entropy of the radiation in the model matches the result read from the quantum extremal surface formula and agrees with the Page curve. Furthermore, we study the reflected entropies between different regions, including the one between the black holes on different sides, the one between the radiations distributed symmetrically but disconnectedly, and the one between the black hole and the radiation on single side. Our results are consistent with the existing ones based on the effective descriptions.

hep-th↗

arXiv2025

Restricted Boltzmann machine as a probabilistic Enigma

We theoretically propose a symmetric encryption scheme based on Restricted Boltzmann Machines that functions as a probabilistic Enigma device, encoding information in the marginal distributions of visible states while utilizing bias permutations as cryptographic keys. Theoretical analysis reveals significant advantages including factorial key space growth through permutation matrices, excellent diffusion properties, and computational complexity rooted in sharp P-complete problems that resist quantum attacks. Compatible with emerging probabilistic computing hardware, the scheme establishes an asymmetric computational barrier where legitimate users decrypt efficiently while adversaries face exponential costs. This framework unlocks probabilistic computers' potential for cryptographic systems, offering an emerging encryption paradigm between classical and quantum regimes for post-quantum security.

cond-mat.stat-mech↗

arXiv2025

KAT-V1: Kwai-AutoThink Technical Report

We present Kwaipilot-AutoThink (KAT), an open-source 40B large language model developed to address the overthinking problem in reasoning-intensive tasks, where an automatic thinking training paradigm is proposed to dynamically switch between reasoning and non-reasoning modes based on task complexity. Specifically, first, we construct the dual-regime dataset based on a novel tagging pipeline and a multi-agent synthesis strategy, and then we apply Multi-Token Prediction (MTP)-enhanced knowledge distillation, enabling efficient and fine-grained reasoning transfer with minimal pretraining cost. Besides, we implement a cold-start initialization strategy that introduces mode-selection priors using majority-vote signals and intent-aware prompting. Finally, we propose Step-SRPO, a reinforcement learning algorithm that incorporates intermediate supervision into the GRPO framework, offering structured guidance over both reasoning-mode selection and response accuracy. Extensive experiments across multiple benchmarks demonstrate that KAT consistently matches or even outperforms current state-of-the-art models, including DeepSeek-R1-0528 and Qwen3-235B-A22B, across a wide range of reasoning-intensive tasks while reducing token usage. Notably, KAT outperforms all open-source models and even surpasses o3-mini on the leakage-controlled LiveCodeBench Pro. Beyond academic evaluation, KAT has been successfully deployed in Kwaipilot (i.e., Kuaishou's internal coding assistant), where it improves real-world development workflows with high accuracy, efficiency, and controllable reasoning behaviors. Moreover, we are actively training a 200B Mixture-of-Experts (MoE) model with 40B active parameters, and early results already show significant gains, further demonstrating the scalability of the AutoThink paradigm.

cs.CL↗

arXiv2025

OASIS: Order-Augmented Strategy for Improved Code Search

Code embeddings capture the semantic representations of code and are crucial for various code-related large language model (LLM) applications, such as code search. Previous training primarily relies on optimizing the InfoNCE loss by comparing positive natural language (NL)-code pairs with in-batch negatives. However, due to the sparse nature of code contexts, training solely by comparing the major differences between positive and negative pairs may fail to capture deeper semantic nuances. To address this issue, we propose a novel order-augmented strategy for improved code search (OASIS). It leverages order-based similarity labels to train models to capture subtle differences in similarity among negative pairs. Extensive benchmark evaluations demonstrate that our OASIS model significantly outperforms previous state-of-the-art models focusing solely on major positive-negative differences. It underscores the value of exploiting subtle differences among negative pairs with order labels for effective code embedding training.

cs.CL↗

arXiv2025

SEER: Semantic Enhancement and Emotional Reasoning Network for Multimodal Fake News Detection

Previous studies on multimodal fake news detection mainly focus on the alignment and integration of cross-modal features, as well as the application of text-image consistency. However, they overlook the semantic enhancement effects of large multimodal models and pay little attention to the emotional features of news. In addition, people find that fake news is more inclined to contain negative emotions than real ones. Therefore, we propose a novel Semantic Enhancement and Emotional Reasoning (SEER) Network for multimodal fake news detection. We generate summarized captions for image semantic understanding and utilize the products of large multimodal models for semantic enhancement. Inspired by the perceived relationship between news authenticity and emotional tendencies, we propose an expert emotional reasoning module that simulates real-life scenarios to optimize emotional features and infer the authenticity of news. Extensive experiments on two real-world datasets demonstrate the superiority of our SEER over state-of-the-art baselines.

cs.MM↗

arXiv2025

Small gaps between almost-twin primes

Let $m \in \mathbb{N}$ be large. We show that there exist infinitely many primes $q_{1}< \cdot\cdot\cdot < q_{m+1}$ such that \[ q_{m+1}-q_{1}=O(e^{7.63m}) \] and $q_{j}+2$ has at most \[ \frac{7.36m}{\log 2} + \frac{4\log m}{\log 2} + 21 \] prime factors for each $1 \leq j \leq m+1$. This improves the previous result of Li and Pan, replacing $m^{4}e^{8m}$ by $e^{7.63m}$ and $\frac{16m}{\log 2} + \frac{5\log m}{\log 2} + 37$ by $\frac{7.36m}{\log 2} + \frac{4\log m}{\log 2} + 21$. The main inputs are the Maynard-Tao sieve, a minorant for the indicator function of the primes constructed by Baker and Irving, for which a stronger equidistribution theorem in arithmetic progressions to smooth moduli is applicable, and Tao's approach previously used to estimate $\sum_{x \leq n < 2x} \mathbf{1}_{\mathbb{P}}(n)\mathbf{1}_{\mathbb{P}}(n+12)ω_{n}$, where $\mathbf{1}_{\mathbb{P}}$ stands for the characteristic function of the primes and $ω_{n}$ are multidimensional sieve weights.

math.NT↗

arXiv2025

Microwave Imaging Spectroscopy Diagnosis of the Slow-rise Precursor of a Major Solar Eruption

In this Letter, taking advantage of microwave data from the Expanded Owens Valley Solar Array and extreme-ultraviolet (EUV) data from the Atmospheric Imaging Assembly, we present the first microwave imaging spectroscopy diagnosis for the slow-rise precursor of a major coronal mass ejection (CME) on 2022 March 30. The EUV images reveal that the CME progenitor, appearing as a hot channel above the polarity inversion line, experiences a slow rise and heating before the eruption. The microwave emissions are found to mainly distribute along the hot channel, with high-frequency sources located at the ends of the hot channel and along precursor bright loops underneath the hot channel. The microwave spectroscopic analysis suggests that microwave emissions in the precursor phase are dominated by thermal emission, largely different from the main phase when a significant non-thermal component is present. These results support the scenario that the precursor reconnection, seeming to be moderate compared with the flare reconnection during the main phase, drives the build-up, heating, and slow rise of CME progenitors toward the explosive eruption.

astro-ph.SR↗