arXiv subjects

Xinyuan Huang

Publications and source records attributed to Xinyuan Huang.

10 recordsLinked to original sources

arXiv2026

MindMemOS: A Portable and Self-Evolving Memory Operating Layer for AI Agents

Memory is a core component of AI agents, enabling them to accumulate experience, maintain personalization, and adapt over long-term interactions. However, existing memory systems often remain fixed after development, limiting their ability to adapt their memory models, organization strategies, and procedural knowledge through continued use. We present MindMemOS, a portable and self-evolving memory operating layer that organizes open-world information using a unified entity property timestructure. MindMemOS supports scenario-adaptive memory modeling, higher-order pattern discovery, autonomous memory refinement, and continuous skill evolution. Its MindMemEvolve algorithm employs validation-driven evolutionary search to optimize memory schemas for target scenarios, whiledreaming consolidates accumulated memories by merging redundant records and resolving conflicts. In addition, implicit corrective feedback serves as a human-in-the-loop signal for identifying and revising potentially inaccurate or misaligned memories. Its MindSkillEvolve algorithm further transforms agent execution trajectories into reusable and progressively refined skills. MindMemOS achieves 94.03% accuracy on LOCOMO and 70.63% on PersonaMem. MindSkillEvolve improves SpreadsheetBench success by 9.2 percentage points over the initial-skill baseline.

cs.AI↗

arXiv2025

S3D: A Spatial Steerable Surgical Drilling Framework for Robotic Spinal Fixation Procedures

In this paper, we introduce S3D: A Spatial Steerable Surgical Drilling Framework for Robotic Spinal Fixation Procedures. S3D is designed to enable realistic steerable drilling while accounting for the anatomical constraints associated with vertebral access in spinal fixation (SF) procedures. To achieve this, we first enhanced our previously designed concentric tube Steerable Drilling Robot (CT-SDR) to facilitate steerable drilling across all vertebral levels of the spinal column. Additionally, we propose a four-Phase calibration, registration, and navigation procedure to perform realistic SF procedures on a spine holder phantom by integrating the CT-SDR with a seven-degree-of-freedom robotic manipulator. The functionality of this framework is validated through planar and out-of-plane steerable drilling experiments in vertebral phantoms.

cs.RO↗

arXiv2025

Federated Neural Architecture Search with Model-Agnostic Meta Learning

Federated Learning (FL) often struggles with data heterogeneity due to the naturally uneven distribution of user data across devices. Federated Neural Architecture Search (NAS) enables collaborative search for optimal model architectures tailored to heterogeneous data to achieve higher accuracy. However, this process is time-consuming due to extensive search space and retraining. To overcome this, we introduce FedMetaNAS, a framework that integrates meta-learning with NAS within the FL context to expedite the architecture search by pruning the search space and eliminating the retraining stage. Our approach first utilizes the Gumbel-Softmax reparameterization to facilitate relaxation of the mixed operations in the search space. We then refine the local search process by incorporating Model-Agnostic Meta-Learning, where a task-specific learner adapts both weights and architecture parameters (alphas) for individual tasks, while a meta learner adjusts the overall model weights and alphas based on the gradient information from task learners. Following the meta-update, we propose soft pruning using the same trick on search space to gradually sparsify the architecture, ensuring that the performance of the chosen architecture remains robust after pruning which allows for immediate use of the model without retraining. Experimental evaluations demonstrate that FedMetaNAS significantly accelerates the search process by more than 50\% with higher accuracy compared to FedNAS.

cs.LG↗

arXiv2024

High-temperature ferromagnetism and ferroelasticity in ultraflexible atomically thin square-shaped lattices

The coexistence of high-temperature intrinsic ferromagnetic ordering, large magnetic anisotropy, along with novel mechanical properties such as ferroelasticity and flexibility, in experimental feasible two-dimensional (2D) crystals is greatly appealing for nanoscale spintronics. However, the progress in identifying such materials is limited. Here, by first-principles calculations, we report the findings of an extraordinary combination of the above qualities for the first time in a new 2D exfoliated FeSi nanosheet in the P4/nmm space group. Due to the strong anion-mediated superexchange interaction, the monolayer FeSi (ML-FeSi) exhibits a Curie temperature Tc as high as 830 K, surpassing the current experimental record (344 K for ML-Cr3Te4). Furthermore, including FeSi, such isostructural lattices all demonstrate exceptional softness, as evidenced by their ultra-low in-plane stiffness. Remarkably, the transition metal atom and square-shaped crystal form work together to give this family of ML materials unique properties that can transition from Ising-like 2D ferromagnets in FeSi, MnP, MnAs, CrP, FeI, and VAs to 2D-XY ones in CrAs, VP, and multiferroic MnGe and TiTe. Overall, our work highlights such 2D lattices as promising candidates in emerging multifunctional device applications and nontrivial topological spintronics.

cond-mat.mes-hall↗

arXiv2022

Independent SE(3)-Equivariant Models for End-to-End Rigid Protein Docking

Protein complex formation is a central problem in biology, being involved in most of the cell's processes, and essential for applications, e.g. drug design or protein engineering. We tackle rigid body protein-protein docking, i.e., computationally predicting the 3D structure of a protein-protein complex from the individual unbound structures, assuming no conformational change within the proteins happens during binding. We design a novel pairwise-independent SE(3)-equivariant graph matching network to predict the rotation and translation to place one of the proteins at the right docked position relative to the second protein. We mathematically guarantee a basic principle: the predicted complex is always identical regardless of the initial locations and orientations of the two structures. Our model, named EquiDock, approximates the binding pockets and predicts the docking poses using keypoint matching and alignment, achieved through optimal transport and a differentiable Kabsch algorithm. Empirically, we achieve significant running time improvements and often outperform existing docking software despite not relying on heavy candidate sampling, structure refinement, or templates.

cs.AI↗

arXiv2021

MedPerf: Open Benchmarking Platform for Medical Artificial Intelligence using Federated Evaluation

Medical AI has tremendous potential to advance healthcare by supporting the evidence-based practice of medicine, personalizing patient treatment, reducing costs, and improving provider and patient experience. We argue that unlocking this potential requires a systematic way to measure the performance of medical AI models on large-scale heterogeneous data. To meet this need, we are building MedPerf, an open framework for benchmarking machine learning in the medical domain. MedPerf will enable federated evaluation in which models are securely distributed to different facilities for evaluation, thereby empowering healthcare organizations to assess and verify the performance of AI models in an efficient and human-supervised process, while prioritizing privacy. We describe the current challenges healthcare and AI communities face, the need for an open platform, the design philosophy of MedPerf, its current implementation status, and our roadmap. We call for researchers and organizations to join us in creating the MedPerf open benchmarking platform.

cs.LG↗

arXiv2021

Benchmarking TinyML Systems: Challenges and Direction

Recent advancements in ultra-low-power machine learning (TinyML) hardware promises to unlock an entirely new class of smart applications. However, continued progress is limited by the lack of a widely accepted benchmark for these systems. Benchmarking allows us to measure and thereby systematically compare, evaluate, and improve the performance of systems and is therefore fundamental to a field reaching maturity. In this position paper, we present the current landscape of TinyML and discuss the challenges and direction towards developing a fair and useful hardware benchmark for TinyML workloads. Furthermore, we present our four benchmarks and discuss our selection methodology. Our viewpoints reflect the collective thoughts of the TinyMLPerf working group that is comprised of over 30 organizations.

cs.PF↗

arXiv2020

MLPerf Training Benchmark

Machine learning (ML) needs industry-standard performance benchmarks to support design and competitive evaluation of the many emerging software and hardware solutions for ML. But ML training presents three unique benchmarking challenges absent from other domains: optimizations that improve training throughput can increase the time to solution, training is stochastic and time to solution exhibits high variance, and software and hardware systems are so diverse that fair benchmarking with the same binary, code, and even hyperparameters is difficult. We therefore present MLPerf, an ML benchmark that overcomes these challenges. Our analysis quantitatively evaluates MLPerf's efficacy at driving performance and scalability improvements across two rounds of results from multiple vendors.

cs.LG↗

arXiv2018

Tiered Object Storage using Persistent Memory

Most data intensive applications often access only a few fields of the objects they are operating on. Since NVM provides fast, byte-addressable access to durable memory, it is possible to access various fields of an object stored in NVM directly without incurring any serialization and deserialization cost. This paper proposes a novel tiered object storage model that modifies a data structure such that only a chosen subset of fields of the data structure are stored in NVM, while the remaining fields are stored in a cheaper (and a traditional) storage layer such as HDDs/SSDs. We introduce a novel linear-programming based optimization framework for deciding the field placement. Our proof of concept demonstrates that a tiered object storage model improves the execution time of standard operations by up to 50\% by avoiding the cost of serialization/deserialization and by reducing the memory footprint of operations.

cs.DC↗

arXiv2010

Preliminary Functional-Structural Modeling on Poplar (Salicaceae)

Poplar is one of the best fast-growing trees in the world, widely used for windbreak and wood product. Although architecture of poplar has direct impact on its applications, it has not been descried in previous poplar models, probably because of the difficulties raised by measurement, data processing and parameterization. In this paper, the functional-structural model GreenLab is calibrated by using poplar data of 3, 4, 5, 6 years old. The data was acquired by simplifying measurement. The architecture was also simplified by classifying the branches into several types (physiological age) using clustering analysis, which decrease the number of parameters. By multi-fitting the sampled data of each tree, the model parameters were identified and the plant architectures at different tree ages were simulated.

math.DS↗