arXiv subjects

Xiao Ma

Publications and source records attributed to Xiao Ma.

At least 55 records · Page 3Linked to original sources

arXiv2025

Human-assisted Robotic Policy Refinement via Action Preference Optimization

Establishing a reliable and iteratively refined robotic system is essential for deploying real-world applications. While Vision-Language-Action (VLA) models are widely recognized as the foundation model for such robotic deployment, their reliance on offline expert demonstrations critically limits their capacity for post-deployment refinement. To mitigate this limitation, we introduce Action Preference Optimization (APO), a method designed to refine VLA models by human-assisted preference alignment gathered through interaction with environments. This method begins with a human-robot collaboration framework for reliable failure correction and interaction trajectory collection through human intervention. However, directly leveraging these interaction trajectories for preference optimization is non-trivial due to the challenges of irreversible robotic actions and token distribution mismatch. To solve this, APO proposes an adaptive reweighting algorithm with binary desirability signals derived from interaction, empowering VLA models effectively suppress failure-prone actions while enhancing corrective action adaptation. Ultimately, APO equips VLA models with the crucial capability to learn from failure, paving the way for their iterative refinement and reliable deployment in dynamic environments. The experiments conducted in simulation and real-world scenarios prove superior generalization and robustness of our human-assisted framework across a variety of manipulation tasks. We believe this work could bring insights for efficient and stable optimization of VLA models through human-robot collaboration. The code and dataset are released at https://github.com/GeWu-Lab/Action-Preference-Optimization

cs.RO↗

arXiv2025

MIREDO: MIP-Driven Resource-Efficient Dataflow Optimization for Computing-in-Memory Accelerator

Computing-in-Memory (CIM) architectures have emerged as a promising solution for accelerating Deep Neural Networks (DNNs) by mitigating data movement bottlenecks. However, realizing the potential of CIM requires specialized dataflow optimizations, which are challenged by an expansive design space and strict architectural constraints. Existing optimization approaches often fail to fully exploit CIM accelerators, leading to noticeable gaps between theoretical and actual system-level efficiency. To address these limitations, we propose the MIREDO framework, which formulates dataflow optimization as a Mixed-Integer Programming (MIP) problem. MIREDO introduces a hierarchical hardware abstraction coupled with an analytical latency model designed to accurately reflect the complex data transfer behaviors within CIM systems. By jointly modeling workload characteristics, dataflow strategies, and CIM-specific constraints, MIREDO systematically navigates the vast design space to determine the optimal dataflow configurations. Evaluation results demonstrate that MIREDO significantly enhances performance, achieving up to $3.2\times$ improvement across various DNN models and hardware setups.

cs.AR↗

arXiv2025

BridgeVLA: Input-Output Alignment for Efficient 3D Manipulation Learning with Vision-Language Models

Recently, leveraging pre-trained vision-language models (VLMs) for building vision-language-action (VLA) models has emerged as a promising approach to effective robot manipulation learning. However, only few methods incorporate 3D signals into VLMs for action prediction, and they do not fully leverage the spatial structure inherent in 3D data, leading to low sample efficiency. In this paper, we introduce BridgeVLA, a novel 3D VLA model that (1) projects 3D inputs to multiple 2D images, ensuring input alignment with the VLM backbone, and (2) utilizes 2D heatmaps for action prediction, unifying the input and output spaces within a consistent 2D image space. In addition, we propose a scalable pre-training method that equips the VLM backbone with the capability to predict 2D heatmaps before downstream policy learning. Extensive experiments show the proposed method is able to learn 3D manipulation efficiently and effectively. BridgeVLA outperforms state-of-the-art baseline methods across three simulation benchmarks. In RLBench, it improves the average success rate from 81.4% to 88.2%. In COLOSSEUM, it demonstrates significantly better performance in challenging generalization settings, boosting the average success rate from 56.7% to 64.0%. In GemBench, it surpasses all the comparing baseline methods in terms of average success rate. In real-robot experiments, BridgeVLA outperforms a state-of-the-art baseline method by 32% on average. It generalizes robustly in multiple out-of-distribution settings, including visual disturbances and unseen instructions. Remarkably, it is able to achieve a success rate of 96.8% on 10+ tasks with only 3 trajectories per task, highlighting its extraordinary sample efficiency. Project Website:https://bridgevla.github.io/

cs.RO↗

arXiv2025

High-resolution velocity model estimation with neural operator and the time-shift imaging condition

Extracting subsurface velocity information from seismic data is mainly an undetermined problem that requires injecting a priori information to constrain the inversion process. Machine learning has offered a platform to do so through the training process, as we formulate our training dataset to inject as much prior knowledge as possible in the trained ML model. Here, we use a neural-operator-based framework for high-resolution seismic velocity model building, which integrates Neural Operators with time-lag reverse time migration imaging. Unlike conventional full waveform inversion (FWI) methods that rely on iterative forward and adjoint-state computations, our approach learns a direct mapping from initial velocity models and extended seismic images to high-resolution velocity estimates through supervised learning. The network architecture enables mesh-independent generalization by learning mappings in infinite-dimensional function spaces, while synthetic velocity models and time-lag reverse time migration (RTM) images provide complementary high-frequency information critical for recovering mid- and high-wavenumber velocity components. Synthetic experiments demonstrate that the proposed method accurately reconstructs fine-scale structures and complex geologies, including out-of-distribution features such as salt bodies. Applications to real seismic field data acquired offshore Australia further validate the method's robustness and resolution capability. The predicted models show enhanced structural details and improved consistency with well-log data, outperforming traditional multi-scale FWI in both accuracy and computational efficiency. The entire prediction process is completed within seconds, making the proposed approach highly suitable for rapid and scalable velocity model building in practical exploration scenarios.

physics.geo-ph↗

arXiv2025

CODA: Repurposing Continuous VAEs for Discrete Tokenization

Discrete visual tokenizers transform images into a sequence of tokens, enabling token-based visual generation akin to language models. However, this process is inherently challenging, as it requires both compressing visual signals into a compact representation and discretizing them into a fixed set of codes. Traditional discrete tokenizers typically learn the two tasks jointly, often leading to unstable training, low codebook utilization, and limited reconstruction quality. In this paper, we introduce \textbf{CODA}(\textbf{CO}ntinuous-to-\textbf{D}iscrete \textbf{A}daptation), a framework that decouples compression and discretization. Instead of training discrete tokenizers from scratch, CODA adapts off-the-shelf continuous VAEs -- already optimized for perceptual compression -- into discrete tokenizers via a carefully designed discretization process. By primarily focusing on discretization, CODA ensures stable and efficient training while retaining the strong visual fidelity of continuous VAEs. Empirically, with $\mathbf{6 \times}$ less training budget than standard VQGAN, our approach achieves a remarkable codebook utilization of 100% and notable reconstruction FID (rFID) of $\mathbf{0.43}$ and $\mathbf{1.34}$ for $8 \times$ and $16 \times$ compression on ImageNet 256$\times$ 256 benchmark.

cs.CV↗

arXiv2025

ByteWrist: A Parallel Robotic Wrist Enabling Flexible and Anthropomorphic Motion for Confined Spaces

This paper introduces ByteWrist, a novel highly-flexible and anthropomorphic parallel wrist for robotic manipulation. ByteWrist addresses the critical limitations of existing serial and parallel wrists in narrow-space operations through a compact three-stage parallel drive mechanism integrated with arc-shaped end linkages. The design achieves precise RPY (Roll-Pitch-Yaw) motion while maintaining exceptional compactness, making it particularly suitable for complex unstructured environments such as home services, medical assistance, and precision assembly. The key innovations include: (1) a nested three-stage motor-driven linkages that minimize volume while enabling independent multi-DOF control, (2) arc-shaped end linkages that optimize force transmission and expand motion range, and (3) a central supporting ball functioning as a spherical joint that enhances structural stiffness without compromising flexibility. Meanwhile, we present comprehensive kinematic modeling including forward / inverse kinematics and a numerical Jacobian solution for precise control. Empirically, we observe ByteWrist demonstrates strong performance in narrow-space maneuverability and dual-arm cooperative manipulation tasks, outperforming Kinova-based systems. Results indicate significant improvements in compactness, efficiency, and stiffness compared to traditional designs, establishing ByteWrist as a promising solution for next-generation robotic manipulation in constrained environments.

cs.RO↗

arXiv2025

Velocity model building from seismic images using a Convolutional Neural Operator

The success of building a high-resolution velocity model using machine learning is hampered by generalization limitations that often limit the success of the approach on field data. This is especially true when relying on neural operators for the mapping. Thus, we propose a novel inversion framework that relies on learning to map the velocity model to a seismic image using a Convolutional Neural Operator (CNO), and then we use optimization to invert for the velocity that matches the image. The key to the success of our network is that we use the initial and true velocity models as input in the training, then we invert for the true velocity starting from the initial velocity at inference. Specifically, we first train a neural operator to accurately learn the forward mapping from seismic velocity models to RTM images, using synthetic datasets that include high-frequency structural information. Once trained, the neural operator is embedded into an inversion loop, where its differentiable nature enables efficient gradient computation via automatic differentiation. This allows us to progressively inject high-wavenumber information from RTM images into the background velocity model, thereby improving resolution without the need for traditional adjoint-state solvers. The proposed framework is validated on both synthetic and field data. Results demonstrate that the neural operator generalizes well to real seismic scenarios, maintains high inversion accuracy, and significantly reduces computational cost. This work highlights the potential of neural operators as flexible and scalable tools for efficient, data-driven seismic imaging and inversion.

physics.geo-ph↗

arXiv2025

GR-3 Technical Report

We report our recent progress towards building generalist robot policies, the development of GR-3. GR-3 is a large-scale vision-language-action (VLA) model. It showcases exceptional capabilities in generalizing to novel objects, environments, and instructions involving abstract concepts. Furthermore, it can be efficiently fine-tuned with minimal human trajectory data, enabling rapid and cost-effective adaptation to new settings. GR-3 also excels in handling long-horizon and dexterous tasks, including those requiring bi-manual manipulation and mobile movement, showcasing robust and reliable performance. These capabilities are achieved through a multi-faceted training recipe that includes co-training with web-scale vision-language data, efficient fine-tuning from human trajectory data collected via VR devices, and effective imitation learning with robot trajectory data. In addition, we introduce ByteMini, a versatile bi-manual mobile robot designed with exceptional flexibility and reliability, capable of accomplishing a wide range of tasks when integrated with GR-3. Through extensive real-world experiments, we show GR-3 surpasses the state-of-the-art baseline method, $π_0$, on a wide variety of challenging tasks. We hope GR-3 can serve as a step towards building generalist robots capable of assisting humans in daily life.

cs.RO↗

arXiv2025

An effective physics-informed neural operator framework for predicting wavefields

Solving the wave equation is fundamental for geophysical applications. However, numerical solutions of the Helmholtz equation face significant computational and memory challenges. Therefore, we introduce a physics-informed convolutional neural operator (PICNO) to solve the Helmholtz equation efficiently. The PICNO takes both the background wavefield corresponding to a homogeneous medium and the velocity model as input function space, generating the scattered wavefield as the output function space. Our workflow integrates PDE constraints directly into the training process, enabling the neural operator to not only fit the available data but also capture the underlying physics governing wave phenomena. PICNO allows for high-resolution reasonably accurate predictions even with limited training samples, and it demonstrates significant improvements over a purely data-driven convolutional neural operator (CNO), particularly in predicting high-frequency wavefields. These features and improvements are important for waveform inversion down the road.

physics.geo-ph↗

arXiv2025

Long time derivation of the Boltzmann equation from hard sphere dynamics

We provide a rigorous derivation of Boltzmann's kinetic equation from the hard sphere system for rarefied gas, which is valid for arbitrarily long times, as long as the solution to the Boltzmann equation exists. This extends Lanford's landmark theorem (1975), which justifies this derivation for a sufficiently short time. In a companion paper (arXiv:2503.01800), we connect this derivation to existing literature on hydrodynamic limits. This completes the resolution of Hilbert's Sixth Problem pertaining to the derivation of fluid equations from Newton's laws, in the case of a rarefied, hard sphere gas. The general strategy follows the paradigm introduced by the first two authors for the long-time derivation of the wave kinetic equation in wave turbulence theory. This is based on propagating a long-time cumulant ansatz, which keeps memory of the full collision history of the relevant particles, by an important partial time expansion. The heart of the matter is proving the smallness of these cumulants in $L^1$, which can be reduced to combinatorial properties for the associated diagrams which we call molecules. These properties are then proved by devising an elaborate cutting algorithm, which is a major novelty of this work.

math.AP↗

arXiv2025

Hyperuniformity at the Absorbing State Transition: Perturbative RG for Random Organization

Hyperuniformity, where the static structure factor obeys $S(q)\sim q^ς$ with $ς> 0$, emerges at criticality in systems having multiple, symmetry-unrelated, absorbing states. Important examples arise in periodically sheared suspensions and amorphous solids; these lie in the random organisation (RO) universality class, for which analytic results for $ς$ are lacking. Here, using Doi-Peliti field theory and perturbative RG about a Gaussian model, we find $ς= 0^+$ and $ς= 2ε/9 + O(ε^2)$ in dimension $d>d_c=4$ and $d=4-ε$ respectively. Our calculations assume that renormalizability is sustained via a certain pattern of cancellation of strongly divergent terms. These cancellations allow the upper critical dimension to remain $d_c = 4$, as is known for RO, while generic perturbations (e.g., those violating particle conservation) would typically flow to a fixed point with $d_c=6$. The assumed cancellation pattern is closely reminiscent of a long-established one near the tricritical Ising fixed point. (This has $d_c=3$, although generic perturbations flow towards the Wilson-Fisher fixed point with $d_c = 4$.) We show how hyperuniformity in RO emerges from anticorrelation of strongly fluctuating active and passive densities. Our calculations also yield the remaining exponents to order $ε$, surprisingly without recourse to functional RG. These exponents coincide as expected with the Conserved Directed Percolation (C-DP) class which also contains the Manna Model and the quenched Edwards-Wilkinson (q-EW) model. Importantly however, our $ς$ differs from one found via a mapping to q-EW. That mapping neglects a conserved noise in the RO action, which we argue to be dangerously irrelevant. Thus, although other exponents are common to both, the RO and C-DP universality classes have different exponents for hyperuniformity.

cond-mat.stat-mech↗

arXiv2025

Evaporation-induced freezing dynamics of droplets levitated in acoustic field

This paper presents the evaporation-induced freezing dynamics of pure cyclohexane droplets levitated via acoustic levitation. Acoustic levitation has attracted considerable attention across various fields owing to its potential to create lab-in-a-drop systems. While droplet evaporation is a fundamental physicochemical process in such a platform, the freezing of droplets induced by evaporation has been sparsely explored experimentally. For pure cyclohexane, the rapid evaporation of levitated droplets initiated freezing at the droplet surface. To better understand this evaporation-induced freezing process, the evaporation behavior of the levitated cyclohexane droplets was visualized and quantified using a high-speed camera and an infrared camera. According to the obtained experimental data, the evaporative heat transfer characteristics of the droplets were identified with theoretical models. Using the derived heat transfer coefficient, a mathematical prediction method for the onset of freezing was proposed and validated with the experimental data. These experimental findings offer valuable insights into the phase transition process and its potential physicochemical applications in a containerless environment.

physics.flu-dyn↗

arXiv2025

Narrow beam and low-sidelobe two-dimensional beam steering on thin-film lithium niobate optical phased array

Optical beam steering has become indispensable in free-space optical communications, light detection and ranging (LiDAR), mapping, and projection. Optical phased array (OPA) leads this field, yet conventional versions still suffer from a narrow steering field of view (FOV), insufficient sidelobe suppression, and limited angular resolution. Thin-film lithium niobate (LN), with its strong Pockels electro-optic (EO) effect, offers a powerful integrated-photonics platform to overcome these limitations. Here we present a two-dimensional (2D) EO-steered OPA based on a non-uniformly spaced X-cut thin-film LN ridge-waveguide array. A superlattice ridge design suppresses optical crosstalk to -20 dB, enabling low-sidelobe far-field radiation. Using particle swarm optimization (PSO) method, we transform a uniformly spaced array into an optimized non-uniform design, largely improving angular resolution while maintaining sidelobe suppression. When combined with a single-radiating trapezoidal end-fire emitter incorporating an etched grating, the device produces a main-lobe beam width of 0.99 degree*0.63 degree from an aperture of only 140 um*250 um, achieving a wide 2D steering range of 47 degree*9.36 degree with a 20 dB sidelobe-suppression ratio. These results highlight thin-film LN OPA as a compelling route toward heterogeneous, compact, and high-performance EO beam-steering modules and ultra-miniaturized optical modulators.

physics.optics↗

arXiv2025

Flow-Based Policy for Online Reinforcement Learning

We present \textbf{FlowRL}, a novel framework for online reinforcement learning that integrates flow-based policy representation with Wasserstein-2-regularized optimization. We argue that in addition to training signals, enhancing the expressiveness of the policy class is crucial for the performance gains in RL. Flow-based generative models offer such potential, excelling at capturing complex, multimodal action distributions. However, their direct application in online RL is challenging due to a fundamental objective mismatch: standard flow training optimizes for static data imitation, while RL requires value-based policy optimization through a dynamic buffer, leading to difficult optimization landscapes. FlowRL first models policies via a state-dependent velocity field, generating actions through deterministic ODE integration from noise. We derive a constrained policy search objective that jointly maximizes Q through the flow policy while bounding the Wasserstein-2 distance to a behavior-optimal policy implicitly derived from the replay buffer. This formulation effectively aligns the flow optimization with the RL objective, enabling efficient and value-aware policy learning despite the complexity of the policy class. Empirical evaluations on DMControl and Humanoidbench demonstrate that FlowRL achieves competitive performance in online reinforcement learning benchmarks.

cs.LG↗

arXiv2025

Coding Theorem for Generalized Reed-Solomon Codes

In this paper, we prove that the sub-field images of generalized Reed-Solomon (RS) codes can achieve the symmetric capacity of p-ary memoryless channels. Unlike the totally random linear code ensemble, as a class of maximum distance separable (MDS) codes, the generalized RS code ensemble lacks the pair-wise independence among codewords and has non-identical distributions of nonzero codewords. To prove the coding theorem for the p-ary images of generalized RS codes, we analyze the exponential upper bound on the error probability of the generalized RS code in terms of its spectrum using random coding techniques. In the finite-length region, we present an ML decoding algorithm for the generalized RS codes over the binary erasure channels (BECs). In particular, the algebraic structure of the generalized RS codes allows us to implement the parallel Lagrange interpolation to derive an ordered systematic matrix. Subsequently, we can reconstruct the ML codeword through a change of basis, accelerating the conventional Gaussian elimination (GE), as validated in the simulation results. Additionally, we apply this decoding technique to the LC-OSD algorithm over the additive white Gaussian noise (AWGN) channels with binary phase shift keying (BPSK) modulation and three-level pulse amplitude modulation (3PAM). Simulation results show that, in the high-rate region, generalized RS codes defined over fields of characteristic three with 3-PAM perform better than those defined over fields of characteristic two with BPSK.

cs.IT↗

arXiv2025

Automated Review Generation Method Based on Large Language Models

Literature research, vital for scientific work, faces the challenge of surging information volumes exceeding researchers' processing capabilities. We present an automated review generation method based on large language models (LLMs) to overcome efficiency bottlenecks and reduce cognitive load. Our statistically validated evaluation framework demonstrates that the generated reviews match or exceed manual quality, offering broad applicability across research fields without requiring users' domain knowledge. Applied to propane dehydrogenation (PDH) catalysts, our method swiftly analyzed 343 articles, averaging seconds per article per LLM account, producing comprehensive reviews spanning 35 topics, with extended analysis of 1041 articles providing insights into catalysts' properties. Through multi-layered quality control, we effectively mitigated LLMs' hallucinations, with expert verification confirming accuracy and citation integrity while demonstrating hallucination risks reduced to below 0.5\% with 95\% confidence. Released Windows application enables one-click review generation, enhancing research productivity and literature recommendation efficiency while setting the stage for broader scientific explorations.

cs.CL↗

arXiv2025

A Short Proof of Coding Theorems for Reed-Muller Codes Under a Mild Assumption

In this paper, by treating Reed-Muller (RM) codes as a special class of low-density parity-check (LDPC) codes and assuming that sub-blocks of the parity-check matrix are randomly interleaved to each other as Gallager's codes, we present a short proof that RM codes are entropy-achieving as source coding for Bernoulli sources and capacity-achieving as channel coding for binary memoryless symmetric (BMS) channels, also known as memoryless binary-input output-symmetric (BIOS) channels, in terms of bit error rate (BER) under maximum-likelihood (ML) decoding.

cs.IT↗

arXiv2025

Systematic Bernoulli Generator Matrix Codes

This paper is concerned with the systematic Bernoulli generator matrix~(BGM) codes, which have been proved to be capacity-achieving over binary-input output-symmetric~(BIOS) channels in terms of bit-error rate~(BER). We prove that the systematic BGM codes are also capacity-achieving over BIOS channels in terms of frame-error rate (FER). To this end, we present a new framework to prove the coding theorems for binary linear codes. Different from the widely-accepted approach via ensemble enlargement, the proof directly applies to the systematic binary linear codes. The new proof indicates that the pair-wise independence condition is not necessary for proving the binary linear code ensemble to achieve the capacity of the BIOS channel. The Bernoulli parity-check~(BPC) codes, which fall within the framework of the systematic BGM codes with parity-check bits known at the decoder can also be proved to achieve the capacity. The presented framework also reveals a new mechanism pertained to the systematic linear codes that the systematic bits and the corresponding parity-check bits play different roles. Precisely, the noisy systematic bits are used to limit the list size of candidate codewords, while the noisy parity-check bits are used to select from the list the maximum likelihood codeword. For systematic BGM codes with finite length, we derive the lower bounds on the BER and FER, which can be used to predict the error floors. Numerical results show that the systematic BGM codes match well with the derived error floors. The performance in water-fall region can be improved with approaches in statistical physics and the error floors can be significantly improved by implementing the concatenated codes with the systematic BGM codes as the inner codes.

cs.IT↗