arXiv subjects

Yuxiang Yang

Publications and source records attributed to Yuxiang Yang.

At least 37 records · Page 2Linked to original sources

arXiv2026

StepAudio 2.5 Technical Report

Unified audio-language modeling has emerged as a prominent trend in modern speech systems, promising to bring the reasoning capabilities of large language models to auditory tasks. However, existing unified foundations often struggle to match the depth of specialized systems across automatic speech recognition (ASR), text-to-speech synthesis (TTS), and realtime spoken interaction. Bridging this gap remains an open challenge. This report presents StepAudio 2.5, a unified audio-language foundation model that matches or exceeds specialized systems across all three capabilities. Rather than treating these tasks as architecturally distinct, we operate on the premise that once text and audio share a multimodal representational space, task specialization becomes a matter of operational regimes: data construction, optimization targets, and decoding constraints. Guided by this insight, we advance the post-training paradigm from standard supervised learning to task-tailored Reinforcement Learning from Human Feedback (RLHF), using it as the primary mechanism to define complex optimization targets. We leverage this RLHF-centric alignment, alongside specialized decoding, to shape a shared backbone into three distinct operational modes. Concretely, the ASR branch advances transcription efficiency via verifiable multi-token decoding; the TTS branch achieves controllable, expressive synthesis through preference-based RLHF and context-rich supervision; and the Realtime branch realizes low-latency, persona-consistent dialogue via generative reward modeling within an RLHF framework. On standard benchmarks, StepAudio 2.5 achieves state-of-the-art results across ASR, TTS, and Realtime, demonstrating that a singular audio-language foundation can successfully internalize the distinct deployment objectives of speech understanding, generation, and live interaction.

eess.AS↗

arXiv2026

Quantum Mpemba Effect in Non-Equilibrium Quantum Thermometry

The quantum Mpemba effect (QMpE) describes an anomalous thermalization phenomenon in which quantum states initially far from equilibrium can approach thermal equilibrium faster than states that begin closer to it. While this effect has been extensively studied in various frameworks, its practical implications for quantum information processing remain largely unexplored. We investigate the relationship between QMpE and quantum thermometry, focusing on non-equilibrium scenarios where measurements are performed during early-stage thermalization. In a Markovian model, we rigorously prove that the initial states that are optimal for thermometry exhibit QMpE with high probability and thermalize faster than most initial states. Our results reveal a fundamental connection between quantum thermodynamics and thermometry, suggesting that QMpE can be harnessed to enhance temperature estimation with quantum probes.

quant-ph↗

arXiv2026

Invisible Adversaries: A Systematic Study of Session Manipulation Attacks on VPNs

Virtual Private Networks (VPNs) are widely used for censorship evasion and traffic protection. VPN users expect to be provided with adequate security protection, and at the same time not be affected by other users connected to the same VPN server, which can be illustrated as the non-interference property. However, in this paper, we have identified several vulnerabilities that violate this property, specifically within the connection tracking frameworks of VPN servers, stemming from shared resource misuse and insufficient validation of session state transitions. We present three session manipulation attacks targeting TCP and UDP traffic tunneled through VPNs. The attacker who only connects to the same VPN server can launch denial-of-service attacks, hijack TCP connections of other clients, or inject forged DNS responses into their queries. We evaluate these attacks against five popular connection tracking frameworks across different OSes and nine major commercial VPN providers. Experimental results reveal that all frameworks and eight providers are vulnerable to at least one of the attacks. We have responsibly disclosed our findings with countermeasures, resulting in 19 assigned CVEs/CNVDs and acknowledgments from the communities and providers.

cs.CR↗

arXiv2026

Critical dynamics of the directed percolation with Lévy-driven temporally quenched disorder

Quenched disorder in absorbing phase transitions can disrupt the structure and symmetry of reaction-diffusion processes, offering a more accurate mapping to real physical systems. We developed a temporally quenched disorder method in the (1+1)-dimensional direct percolation (DP) model, where the increment of conditional probability is determined by the cumulative distribution function (CDF) of the Lévy distribution. Monte Carlo (MC) simulations reveal that the model has a critical region governing the transition between absorbing and active states, and this region changes as the parameter $β$, which influences distribution properties. Guided by dynamic scaling laws, we observe that significant variations in the Lévy distribution parameter $β$ lead to notable changes in the particle density decay exponent $α$, total particle number exponent $θ$, and spreading exponent $\tilde{z}$. The quenching mechanism we introduced has broad potential applications in various theoretical and experimental studies of absorbing phase transitions.

cond-mat.stat-mech↗

arXiv2026

JOintGS: Joint Optimization of Cameras, Bodies and 3D Gaussians for In-the-Wild Monocular Reconstruction

Reconstructing high-fidelity animatable 3D human avatars from monocular RGB videos remains challenging, particularly in unconstrained in-the-wild scenarios where camera parameters and human poses from off-the-shelf methods (e.g., COLMAP, HMR2.0) are often inaccurate. Splatting (3DGS) advances demonstrate impressive rendering quality and real-time performance, they critically depend on precise camera calibration and pose annotations, limiting their applicability in real-world settings. We present JOintGS, a unified framework that jointly optimizes camera extrinsics, human poses, and 3D Gaussian representations from coarse initialization through a synergistic refinement mechanism. Our key insight is that explicit foreground-background disentanglement enables mutual reinforcement: static background Gaussians anchor camera estimation via multi-view consistency; refined cameras improve human body alignment through accurate temporal correspondence; optimized human poses enhance scene reconstruction by removing dynamic artifacts from static constraints. We further introduce a temporal dynamics module to capture fine-grained pose-dependent deformations and a residual color field to model illumination variations. Extensive experiments on NeuMan and EMDB datasets demonstrate that JOintGS achieves superior reconstruction quality, with 2.1~dB PSNR improvement over state-of-the-art methods on NeuMan dataset, while maintaining real-time rendering. Notably, our method shows significantly enhanced robustness to noisy initialization compared to the baseline.Our source code is available at https://github.com/MiliLab/JOintGS.

cs.CV↗

arXiv2026

STEP3-VL-10B Technical Report

We present STEP3-VL-10B, a lightweight open-source foundation model designed to redefine the trade-off between compact efficiency and frontier-level multimodal intelligence. STEP3-VL-10B is realized through two strategic shifts: first, a unified, fully unfrozen pre-training strategy on 1.2T multimodal tokens that integrates a language-aligned Perception Encoder with a Qwen3-8B decoder to establish intrinsic vision-language synergy; and second, a scaled post-training pipeline featuring over 1k iterations of reinforcement learning. Crucially, we implement Parallel Coordinated Reasoning (PaCoRe) to scale test-time compute, allocating resources to scalable perceptual reasoning that explores and synthesizes diverse visual hypotheses. Consequently, despite its compact 10B footprint, STEP3-VL-10B rivals or surpasses models 10$\times$-20$\times$ larger (e.g., GLM-4.6V-106B, Qwen3-VL-235B) and top-tier proprietary flagships like Gemini 2.5 Pro and Seed-1.5-VL. Delivering best-in-class performance, it records 92.2% on MMBench and 80.11% on MMMU, while excelling in complex reasoning with 94.43% on AIME2025 and 75.95% on MathVision. We release the full model suite to provide the community with a powerful, efficient, and reproducible baseline.

cs.CV↗

arXiv2026

InpaintHuman: Reconstructing Occluded Humans with Multi-Scale UV Mapping and Identity-Preserving Diffusion Inpainting

Reconstructing complete and animatable 3D human avatars from monocular videos remains challenging, particularly under severe occlusions. While 3D Gaussian Splatting has enabled photorealistic human rendering, existing methods struggle with incomplete observations, often producing corrupted geometry and temporal inconsistencies. We present InpaintHuman, a novel method for generating high-fidelity, complete, and animatable avatars from occluded monocular videos. Our approach introduces two key innovations: (i) a multi-scale UV-parameterized representation with hierarchical coarse-to-fine feature interpolation, enabling robust reconstruction of occluded regions while preserving geometric details; and (ii) an identity-preserving diffusion inpainting module that integrates textual inversion with semantic-conditioned guidance for subject-specific, temporally coherent completion. Unlike SDS-based methods, our approach employs direct pixel-level supervision to ensure identity fidelity. Experiments on synthetic benchmarks (PeopleSnapshot, ZJU-MoCap) and real-world scenarios (OcMotion) demonstrate competitive performance with consistent improvements in reconstruction quality across diverse poses and viewpoints.

cs.CV↗

arXiv2025

Recovering optimal precision in quantum sensing with time domain imperfections

Quantum control plays a crucial role in enhancing precision scaling for quantum sensing. However, most existing protocols require perfect control, even though real-world devices inevitably have control imperfections. Here, we consider a fundamental setting of quantum sensing with time domain imperfections, where the duration of control pulses and the interrogation time are all subject to uncertainty. Under this scenario, we investigate the task of frequency estimation in the presence of a non-Markovian environment. We design a control strategy and prove that it outperforms any control-free strategies, recovering the optimal Heisenberg limit up to a small error term that is intrinsic to this model. We further demonstrate the advantage of our control strategy via experiments on a nuclear magnetic resonance (NMR) platform. Our finding confirms that the advantage of quantum control in quantum sensing persists even in the presence of imperfections.

quant-ph↗

arXiv2025

Optimal quantum metrology under energy constraints

The traditional framework of quantum metrology commonly assumes unlimited access to resources, overlooking resource constraints in realistic scenarios. As such, the optimal strategies therein can be infeasible in practice. Here, we investigate quantum metrology where the total energy consumption of the probe state preparation, intermediate control operations, and the final measurement is subject to a constraint. We establish a comprehensive theoretical framework for characterizing energy-constrained multi-step quantum processes, based on which we develop a general optimization method for energy-constrained quantum metrology that determines both the optimal precision and the corresponding strategy. Using the method, we determine the ultimate precision limit of energy-constrained phase estimation and identify a novel advantage of quantum superpositions of causal orders in enhancing the energy efficiency of adaptive quantum estimation.

quant-ph↗

arXiv2025

Uncovering Gaps Between RFC Updates and TCP/IP Implementations: LLM-Facilitated Differential Checks on Intermediate Representations

As the core of the Internet infrastructure, the TCP/IP protocol stack undertakes the task of network data transmission. However, due to the complexity of the protocol and the uncertainty of cross-layer interaction, there are often inconsistencies between the implementation of the protocol stack code and the RFC standard. This inconsistency may not only lead to differences in protocol functions but also cause serious security vulnerabilities. At present, with the continuous expansion of protocol stack functions and the rapid iteration of RFC documents, it is increasingly important to detect and fix these inconsistencies. With the rise of large language models, researchers have begun to explore how to extract protocol specifications from RFC documents through these models, including protocol stack modeling, state machine extraction, text ambiguity analysis, and other related content. However, existing methods rely on predefined patterns or rule-based approaches that fail to generalize across different protocol specifications. Automated and scalable detection of these inconsistencies remains a significant challenge. In this study, we propose an automated analysis framework based on LLM and differential models. By modeling the iterative relationship of the protocol and based on the iterative update relationship of the RFC standard, we perform incremental code function analysis on different versions of kernel code implementations to automatically perform code detection and vulnerability analysis. We conduct extensive evaluations to validate the effectiveness of our framework, demonstrating its effectiveness in identifying potential vulnerabilities caused by RFC code inconsistencies.

cs.CR↗

arXiv2025

BEVTrack: A Simple and Strong Baseline for 3D Single Object Tracking in Bird's-Eye View

3D Single Object Tracking (SOT) is a fundamental task in computer vision and plays a critical role in applications like autonomous driving. However, existing algorithms often involve complex designs and multiple loss functions, making model training and deployment challenging. Furthermore, their reliance on fixed probability distribution assumptions (e.g., Laplacian or Gaussian) hinders their ability to adapt to diverse target characteristics such as varying sizes and motion patterns, ultimately affecting tracking precision and robustness. To address these issues, we propose BEVTrack, a simple yet effective motion-based tracking method. BEVTrack directly estimates object motion in Bird's-Eye View (BEV) using a single regression loss. To enhance accuracy for targets with diverse attributes, it learns adaptive likelihood functions tailored to individual targets, avoiding the limitations of fixed distribution assumptions in previous methods. This approach provides valuable priors for tracking and significantly boosts performance. Comprehensive experiments on KITTI, NuScenes, and Waymo Open Dataset demonstrate that BEVTrack achieves state-of-the-art results while operating at 200 FPS, enabling real-time applicability. The code will be released at https://github.com/xmm-prio/BEVTrack.

cs.CV↗

arXiv2025

Heisenberg-Limited Quantum Metrology without Ancillae

Extensive research has been dedicated to the asymptotic theory of quantum metrology, where the goal is to determine the ultimate precision limit of quantum channel estimation when many accesses to the channel are allowed. The ultimate limit has been well established, but in general a noiseless and controllable ancilla is required for attaining it. Little is known about the metrological performance without noiseless ancillae, which is more relevant in practical circumstances. In this Letter, we present a novel theoretical framework to address this problem, bridging quantum metrology and the asymptotic theory of quantum channels. Leveraging this framework, we prove sufficient conditions for achieving the Heisenberg limit with repeated applications of the channel to be estimated, both with and without applying interleaved unitary control operations. For the latter case, we design an algorithm to identify the control operation explicitly.

quant-ph↗

arXiv2025

FMGS-Avatar: Mesh-Guided 2D Gaussian Splatting with Foundation Model Priors for 3D Monocular Avatar Reconstruction

Reconstructing high-fidelity animatable human avatars from monocular videos remains challenging due to insufficient geometric information in single-view observations. While recent 3D Gaussian Splatting methods have shown promise, they struggle with surface detail preservation due to the free-form nature of 3D Gaussian primitives. To address both the representation limitations and information scarcity, we propose a novel method, \textbf{FMGS-Avatar}, that integrates two key innovations. First, we introduce Mesh-Guided 2D Gaussian Splatting, where 2D Gaussian primitives are attached directly to template mesh faces with constrained position, rotation, and movement, enabling superior surface alignment and geometric detail preservation. Second, we leverage foundation models trained on large-scale datasets, such as Sapiens, to complement the limited visual cues from monocular videos. However, when distilling multi-modal prior knowledge from foundation models, conflicting optimization objectives can emerge as different modalities exhibit distinct parameter sensitivities. We address this through a coordinated training strategy with selective gradient isolation, enabling each loss component to optimize its relevant parameters without interference. Through this combination of enhanced representation and coordinated information distillation, our approach significantly advances 3D monocular human avatar reconstruction. Experimental evaluation demonstrates superior reconstruction quality compared to existing methods, with notable gains in geometric accuracy and appearance fidelity while providing rich semantic information. Additionally, the distilled prior knowledge within a shared canonical space naturally enables spatially and temporally consistent rendering under novel views and poses.

cs.CV↗

arXiv2025

Supervised and unsupervised learning with numerical computation for the Wolfram cellular automata

The local rules of Wolfram cellular automata with one-dimensional three-cell neighborhoods are represented by eight-bit binary that encode deterministic update rules. These automata are widely utilized to investigate self-organization phenomena and the dynamics of complex systems. In this work, we employ numerical simulations and computational methods to investigate the asymptotic density and dynamical evolution mechanisms in Wolfram automata. We apply both supervised and unsupervised learning methods to identify the configurations associated with different Wolfram rules. Furthermore, we explore alternative initial conditions under which certain Wolfram rules generate similar fractal patterns over time, even when starting from a single active site. Our results reveal the relationship between the asymptotic density and the initial density of selected rules. The supervised learning methods effectively identify the configurations of various Wolfram rules, while unsupervised methods like principal component analysis and autoencoders can approximately cluster configurations of different Wolfram rules into distinct groups, yielding results that align well with simulated density outputs.

physics.comp-ph↗

arXiv2025

LocoTouch: Learning Dynamic Quadrupedal Transport with Tactile Sensing

Quadrupedal robots have demonstrated remarkable agility and robustness in traversing complex terrains. However, they struggle with dynamic object interactions, where contact must be precisely sensed and controlled. To bridge this gap, we present LocoTouch, a system that equips quadrupedal robots with tactile sensing to address a particularly challenging task in this category: long-distance transport of unsecured cylindrical objects, which typically requires custom mounting or fastening mechanisms to maintain stability. For efficient large-area tactile sensing, we design a high-density distributed tactile sensor that covers the entire back of the robot. To effectively leverage tactile feedback for robot control, we develop a simulation environment with high-fidelity tactile signals, and train tactile-aware transport policies using a two-stage learning pipeline. Furthermore, we design a novel reward function to promote robust, symmetric, and frequency-adaptive locomotion gaits. After training in simulation, LocoTouch transfers zero-shot to the real world, reliably transporting a wide range of unsecured cylindrical objects with diverse sizes, weights, and surface properties. Moreover, it remains robust over long distances, on uneven terrain, and under severe perturbations.

cs.RO↗

arXiv2025

Step-Audio 2 Technical Report

This paper presents Step-Audio 2, an end-to-end multi-modal large language model designed for industry-strength audio understanding and speech conversation. By integrating a latent audio encoder and reasoning-centric reinforcement learning (RL), Step-Audio 2 achieves promising performance in automatic speech recognition (ASR) and audio understanding. To facilitate genuine end-to-end speech conversation, Step-Audio 2 incorporates the generation of discrete audio tokens into language modeling, significantly enhancing its responsiveness to paralinguistic information such as speaking styles and emotions. To effectively leverage the rich textual and acoustic knowledge in real-world data, Step-Audio 2 integrates retrieval-augmented generation (RAG) and is able to call external tools such as web search to mitigate hallucination and audio search to switch timbres. Trained on millions of hours of speech and audio data, Step-Audio 2 delivers intelligence and expressiveness across diverse conversational scenarios. Evaluation results demonstrate that Step-Audio 2 achieves state-of-the-art performance on various audio understanding and conversational benchmarks compared to other open-source and commercial solutions. Please visit https://github.com/stepfun-ai/Step-Audio2 for more information.

cs.CL↗

arXiv2025

Human2LocoMan: Learning Versatile Quadrupedal Manipulation with Human Pretraining

Quadrupedal robots have demonstrated impressive locomotion capabilities in complex environments, but equipping them with autonomous versatile manipulation skills in a scalable way remains a significant challenge. In this work, we introduce a cross-embodiment imitation learning system for quadrupedal manipulation, leveraging data collected from both humans and LocoMan, a quadruped equipped with multiple manipulation modes. Specifically, we develop a teleoperation and data collection pipeline, which unifies and modularizes the observation and action spaces of the human and the robot. To effectively leverage the collected data, we propose an efficient modularized architecture that supports co-training and pretraining on structured modality-aligned data across different embodiments. Additionally, we construct the first manipulation dataset for the LocoMan robot, covering various household tasks in both unimanual and bimanual modes, supplemented by a corresponding human dataset. We validate our system on six real-world manipulation tasks, where it achieves an average success rate improvement of 41.9% overall and 79.7% under out-of-distribution (OOD) settings compared to the baseline. Pretraining with human data contributes a 38.6% success rate improvement overall and 82.7% under OOD settings, enabling consistently better performance with only half the amount of robot data. Our code, hardware, and data are open-sourced at: https://human2bots.github.io.

cs.RO↗

arXiv2025

Step-Audio-AQAA: a Fully End-to-End Expressive Large Audio Language Model

Large Audio-Language Models (LALMs) have significantly advanced intelligent human-computer interaction, yet their reliance on text-based outputs limits their ability to generate natural speech responses directly, hindering seamless audio interactions. To address this, we introduce Step-Audio-AQAA, a fully end-to-end LALM designed for Audio Query-Audio Answer (AQAA) tasks. The model integrates a dual-codebook audio tokenizer for linguistic and semantic feature extraction, a 130-billion-parameter backbone LLM and a neural vocoder for high-fidelity speech synthesis. Our post-training approach employs interleaved token-output of text and audio to enhance semantic coherence and combines Direct Preference Optimization (DPO) with model merge to improve performance. Evaluations on the StepEval-Audio-360 benchmark demonstrate that Step-Audio-AQAA excels especially in speech control, outperforming the state-of-art LALMs in key areas. This work contributes a promising solution for end-to-end LALMs and highlights the critical role of token-based vocoder in enhancing overall performance for AQAA tasks.

cs.SD↗