arXiv subjects

Yi Xu

Publications and source records attributed to Yi Xu.

At least 163 records · Page 9Linked to original sources

arXiv2025

RLGS: Reinforcement Learning-Based Adaptive Hyperparameter Tuning for Gaussian Splatting

Hyperparameter tuning in 3D Gaussian Splatting (3DGS) is a labor-intensive and expert-driven process, often resulting in inconsistent reconstructions and suboptimal results. We propose RLGS, a plug-and-play reinforcement learning framework for adaptive hyperparameter tuning in 3DGS through lightweight policy modules, dynamically adjusting critical hyperparameters such as learning rates and densification thresholds. The framework is model-agnostic and seamlessly integrates into existing 3DGS pipelines without architectural modifications. We demonstrate its generalization ability across multiple state-of-the-art 3DGS variants, including Taming-3DGS and 3DGS-MCMC, and validate its robustness across diverse datasets. RLGS consistently enhances rendering quality. For example, it improves Taming-3DGS by 0.7dB PSNR on the Tanks and Temple (TNT) dataset, under a fixed Gaussian budget, and continues to yield gains even when baseline performance saturates. Our results suggest that RLGS provides an effective and general solution for automating hyperparameter tuning in 3DGS training, bridging a gap in applying reinforcement learning to 3DGS.

cs.GR↗

arXiv2025

ConformalSAM: Unlocking the Potential of Foundational Segmentation Models in Semi-Supervised Semantic Segmentation with Conformal Prediction

Pixel-level vision tasks, such as semantic segmentation, require extensive and high-quality annotated data, which is costly to obtain. Semi-supervised semantic segmentation (SSSS) has emerged as a solution to alleviate the labeling burden by leveraging both labeled and unlabeled data through self-training techniques. Meanwhile, the advent of foundational segmentation models pre-trained on massive data, has shown the potential to generalize across domains effectively. This work explores whether a foundational segmentation model can address label scarcity in the pixel-level vision task as an annotator for unlabeled images. Specifically, we investigate the efficacy of using SEEM, a Segment Anything Model (SAM) variant fine-tuned for textual input, to generate predictive masks for unlabeled data. To address the shortcomings of using SEEM-generated masks as supervision, we propose ConformalSAM, a novel SSSS framework which first calibrates the foundation model using the target domain's labeled data and then filters out unreliable pixel labels of unlabeled data so that only high-confidence labels are used as supervision. By leveraging conformal prediction (CP) to adapt foundation models to target data through uncertainty calibration, ConformalSAM exploits the strong capability of the foundational segmentation model reliably which benefits the early-stage learning, while a subsequent self-reliance training strategy mitigates overfitting to SEEM-generated masks in the later training stage. Our experiment demonstrates that, on three standard benchmarks of SSSS, ConformalSAM achieves superior performance compared to recent SSSS methods and helps boost the performance of those methods as a plug-in.

cs.CV↗

arXiv2025

Dark-state engineering in Fock-state lattices

Fock-state lattices (FSLs) are becoming an emerging research hotspot in quantum physics, not only because the FSLs provide a new perspective for studying atom-field interactions, but also because they build the connection between quantum optics and condensed matter physics. Owing to the multiple transition paths in the lattices, inherent quantum interference effect exists in these systems, and hence how to find new quantum coherent phenomena and exploit their applications becomes a significant and desired task in this field. In this work, we study the dark-state effect in the FSLs generated by the multimode Jaynes-Cummings (JC) models. By considering the FSLs in certain-excitation-number subspaces, we study the dark states with respect to the states associated with the atomic excited state using the arrowhead-matrix method. We find that there exist dark-state subspaces with the dimensions determined by the number of orthogonal dark states. When the dimension is larger than one, the forms of these dark-state bases are not unique. Further, we obtain the number and form of the orthogonal dark states in the two-, three-, and four-mode JC models. In addition, we find that for a general $N$-mode JC model, there are $C_{N+n-2}^{N-2}$ orthogonal dark states in the $n$-excitation subspace. We also build the relationship between the dark modes and dark states. Our work will pave the way for exploring quantum optical effects and quantum information processing based on the FSLs.

quant-ph↗

arXiv2025

Postive Semidefinite and Sum of Squares Biquadratic Polynomials

Hilbert proved in 1888 that a positive semi-definite (PSD) homogeneous quartic polynomial of three variables always can be expressed as the sum of squares (SOS) of three quadratic polynomials, and a psd homogeneous quartic polynomial of four variables may not be sos. Only after 87 years, in 1975, Choi gave the explicit expression of such a psd-not-sos (PNS) homogeneous quartic polynomial of four variables. An $m \times n$ biquadratic polynomial is a homogeneous quartic polynomial of $m+n$ variables. In this paper, we show that an $m \times n$ biquadratic polynomial can be expressed as a tripartite homogeneous quartic polynomial of $m+n-1$ variables. Therefore, {by Hilbert's theorem}, a $2 \times 2$ PSD biquadratic polynomial can be expressed as the sum of squares of three quadratic polynomials. This improves the result of Calderón in 1973, who proved that a $2 \times 2$ biquadratic polynomial can be expressed as the sum of squares of nine quadratic polynomials. Furthermore, we present a necessary and sufficient condition for an $m \times n$ psd biquadratic polynomial to be sos, and show that if such a polynomial is sos, then its sos rank is at most $mn$. Then we give a constructive proof of the sos form of a $2 \times 2$ psd biquadratic polynomial in three cases.

math.OC↗

arXiv2025

Stochastic Weakly Convex Optimization Under Heavy-Tailed Noises

An increasing number of studies have focused on stochastic first-order methods (SFOMs) under heavy-tailed gradient noises, which have been observed in the training of practical deep learning models. In this paper, we focus on two types of gradient noises: one is sub-Weibull noise, and the other is noise under the assumption that it has a bounded $p$-th central moment ($p$-BCM) with $p\in (1, 2]$. The latter is more challenging due to the occurrence of infinite variance when $p\in (1, 2)$. Under these two gradient noise assumptions, the in-expectation and high-probability convergence of SFOMs have been extensively studied in the contexts of convex optimization and standard smooth optimization. However, for weakly convex objectives-a class that includes all Lipschitz-continuous convex objectives and smooth objectives-our understanding of the in-expectation and high-probability convergence of SFOMs under these two types of noises remains incomplete. We investigate the high-probability convergence of the vanilla stochastic subgradient descent (SsGD) method under sub-Weibull noises, as well as the high-probability and in-expectation convergence of clipped SsGD under the $p$-BCM noises. Both analyses are conducted in the context of weakly convex optimization. For weakly convex objectives that may be non-convex and non-smooth, our results demonstrate that the theoretical dependence of vanilla SsGD on the failure probability and number of iterations under sub-Weibull noises does not degrade compared to the case of smooth objectives. Under $p$-BCM noises, our findings indicate that the non-smoothness and non-convexity of weakly convex objectives do not impact the theoretical dependence of clipped SGD on the failure probability relative to the smooth case; however, the sample complexity we derived is worse than a well-known lower bound for smooth optimization.

math.OC↗

arXiv2025

MoRE: A Mixture of Reflectors Framework for Large Language Model-Based Sequential Recommendation

Large language models (LLMs) have emerged as a cutting-edge approach in sequential recommendation, leveraging historical interactions to model dynamic user preferences. Current methods mainly focus on learning processed recommendation data in the form of sequence-to-sequence text. While effective, they exhibit three key limitations: 1) failing to decouple intra-user explicit features (e.g., product titles) from implicit behavioral patterns (e.g., brand loyalty) within interaction histories; 2) underutilizing cross-user collaborative filtering (CF) signals; and 3) relying on inefficient reflection update strategies. To address this, We propose MoRE (Mixture of REflectors), which introduces three perspective-aware offline reflection processes to address these gaps. This decomposition directly resolves Challenges 1 (explicit/implicit ambiguity) and 2 (CF underutilization). Furthermore, MoRE's meta-reflector employs a self-improving strategy and a dynamic selection mechanism (Challenge 3) to adapt to evolving user preferences. First, two intra-user reflectors decouple explicit and implicit patterns from a user's interaction sequence, mimicking traditional recommender systems' ability to distinguish surface-level and latent preferences. A third cross-user reflector captures CF signals by analyzing user similarity patterns from multiple users' interactions. To optimize reflection quality, MoRE's meta-reflector employs a offline self-improving strategy that evaluates reflection impacts through comparisons of presence/absence and iterative refinement of old/new versions, with a online contextual bandit mechanism dynamically selecting the optimal perspective for recommendation for each user. Code: https://github.com/E-qin/MoRE-Rec.

cs.IR↗

arXiv2025

Simulating Spin Dynamics of Supersolid States in a Quantum Ising Magnet

Motivated by a recent experimental study on the quantum Ising magnet $\text{K}_2\text{Co}(\text{SeO}_3)_2$ that presented spectroscopic evidence of zero-field supersolidity (Chen et al., arXiv:2402.15869), we simulate the excitation spectrum of the corresponding microscopic $XXZ$ model for the compound, using the recently developed excitation ansatz for infinite projected entangled-pair states. We map out the ground state phase diagram and compute the dynamical spin structure factors across a range of magnetic field strengths, focusing especially on the two supersolid phases found near zero and saturation fields. Our simulated excitation spectra for the zero-field supersolid "Y" phase are in excellent agreement with the experimental data - recovering the low-energy branches and integer quantized excited energy levels $ω_n=nJ_{zz}$. Furthermore, we demonstrate the nonlocal multi-spin-flip features for modes at $ω_2$, indicative of their multi-magnon nature. Additionally, we identify characteristics of the high-field supersolid "$Ψ$" phase in the simulated spectra, which should be compared with future experimental results.

cond-mat.str-el↗

arXiv2025

Score-Based Model for Low-Rank Tensor Recovery

Low-rank tensor decompositions (TDs) provide an effective framework for multiway data analysis. Traditional TD methods rely on predefined structural assumptions, such as CP or Tucker decompositions. From a probabilistic perspective, these can be viewed as using Dirac delta distributions to model the relationships between shared factors and the low-rank tensor. However, such prior knowledge is rarely available in practical scenarios, particularly regarding the optimal rank structure and contraction rules. The optimization procedures based on fixed contraction rules are complex, and approximations made during these processes often lead to accuracy loss. To address this issue, we propose a score-based model that eliminates the need for predefined structural or distributional assumptions, enabling the learning of compatibility between tensors and shared factors. Specifically, a neural network is designed to learn the energy function, which is optimized via score matching to capture the gradient of the joint log-probability of tensor entries and shared factors. Our method allows for modeling structures and distributions beyond the Dirac delta assumption. Moreover, integrating the block coordinate descent (BCD) algorithm with the proposed smooth regularization enables the model to perform both tensor completion and denoising. Experimental results demonstrate significant performance improvements across various tensor types, including sparse and continuous-time tensors, as well as visual data.

cs.LG↗

arXiv2025

Similarity = Value? Consultation Value Assessment and Alignment for Personalized Search

Personalized search systems in e-commerce platforms increasingly involve user interactions with AI assistants, where users consult about products, usage scenarios, and more. Leveraging consultation to personalize search services is trending. Existing methods typically rely on semantic similarity to align historical consultations with current queries due to the absence of 'value' labels, but we observe that semantic similarity alone often fails to capture the true value of consultation for personalization. To address this, we propose a consultation value assessment framework that evaluates historical consultations from three novel perspectives: (1) Scenario Scope Value, (2) Posterior Action Value, and (3) Time Decay Value. Based on this, we introduce VAPS, a value-aware personalized search model that selectively incorporates high-value consultations through a consultation-user action interaction module and an explicit objective that aligns consultations with user actions. Experiments on both public and commercial datasets show that VAPS consistently outperforms baselines in both retrieval and ranking tasks.

cs.IR↗

arXiv2025

Collimated Hard X-Rays from Hybrid Laser and Plasma Wakefield Accelerators

We report a synergistic enhancement of betatron radiation based on the hybrid laser and plasma wakefield acceleration scheme. Quasi-phase-stable acceleration in an up-ramp plasma density first generates GeV-energy electron beams that act as a drive beam for PWFA, which then further accelerates the witness beam to GeV energies, enhancing both photon energy and flux. A full width at half maximum divergence $(6.1 \pm 1.9)\times(5.8\pm 1.6) $ mrad$^2$ of betatron radiation, a critical energy of $71 \pm 8$ keV, and an average flux of more than $10^{14}$ photons per steradian above 5 keV were all experimentally obtained thanks to this scheme, which was an order of magnitude higher than the previous reports. Quasi-three-dimensional particle-in-cell simulations were used to model the acceleration and radiation of the electrons in our experimental conditions, establishing a new paradigm for compact collimated hard X-ray sources.

physics.plasm-ph↗

arXiv2025

Snap, Segment, Deploy: A Visual Data and Detection Pipeline for Wearable Industrial Assistants

Industrial assembly tasks increasingly demand rapid adaptation to complex procedures and varied components, yet are often conducted in environments with limited computing, connectivity, and strict privacy requirements. These constraints make conventional cloud-based or fully autonomous solutions impractical for factory deployment. This paper introduces a mobile-device-based assistant system for industrial training and operational support, enabling real-time, semi-hands-free interaction through on-device perception and voice interfaces. The system integrates lightweight object detection, speech recognition, and Retrieval-Augmented Generation (RAG) into a modular on-device pipeline that operates entirely on-device, enabling intuitive support for part handling and procedure understanding without relying on manual supervision or cloud services. To enable scalable training, we adopt an automated data construction pipeline and introduce a two-stage refinement strategy to improve visual robustness under domain shift. Experiments on our generated dataset, i.e., Gear8, demonstrate improved robustness to domain shift and common visual corruptions. A structured user study further confirms its practical viability, with positive user feedback on the clarity of the guidance and the quality of the interaction. These results indicate that our framework offers a deployable solution for real-time, privacy-preserving smart assistance in industrial environments. We will release the Gear8 dataset and source code upon acceptance.

cs.HC↗

arXiv2025

Investigating Non-Transitivity in LLM-as-a-Judge

Automatic evaluation methods based on large language models (LLMs) are emerging as the standard tool for assessing the instruction-following abilities of LLM-based agents. The most common method in this paradigm, pairwise comparisons with a baseline model, critically depends on the assumption of transitive preferences. However, the validity of this assumption remains largely unexplored. In this study, we investigate the presence of non-transitivity within the AlpacaEval framework and analyze its effects on model rankings. We find that LLM judges exhibit non-transitive preferences, leading to rankings that are sensitive to the choice of the baseline model. To mitigate this issue, we show that round-robin tournaments combined with Bradley-Terry models of preference can produce more reliable rankings. Notably, our method increases both the Spearman correlation and the Kendall correlation with Chatbot Arena (95.0% -> 96.4% and 82.1% -> 86.3% respectively). To address the computational cost of round-robin tournaments, we propose Swiss-Wise Iterative Matchmaking (Swim) tournaments, using a dynamic matching strategy to capture the benefits of round-robin tournaments while maintaining computational efficiency.

cs.AI↗

arXiv2025

Adaptive recycled plastic architecture: Vacuum-Sealed Chainmail Structures Through Computational Design

The construction industry is a major consumer of raw materials, accounting for nearly half of global material usage annually, while generating significant waste that poses sustainability challenges. This paper explores the untapped potential of recycled plastics as a primary construction material, leveraging their lightweight, flexible, and customizable properties for advanced applications in modular chainmail systems. Through a computational workflow, the study optimizes the design, testing, and fabrication of vacuum-sealed chainmail structures composed of recycled plastic filaments, demonstrating their adaptability and structural performance for architectural use. Key contributions include a novel methodology for integrating recycled plastic filaments into chainmail geometries, validated through 2D sectional testing, 3D shell structure generation, and physical modeling under vacuum constraints. The research identifies the rectangular chainmail configuration as the most efficient and adaptable, achieving superior deformation capacity, material efficiency, and load-bearing performance. Optimization strategies for temporary structures highlight practical deployment potential, balancing material savings, usable area, and water drainage efficiency. The findings offer a foundation for innovative applications in extreme conditions, including disaster-prone areas, high-altitude environments, underwater platforms, and extraterrestrial habitats. These applications leverage the lightweight, adaptable, and durable properties of recycled plastics and modular chainmail systems, bridging the gap between waste management and high-performance design while addressing unique challenges in harsh and resource-constrained environments.

cs.CE↗

arXiv2025

AutoMixAlign: Adaptive Data Mixing for Multi-Task Preference Optimization in LLMs

When aligning large language models (LLMs), their performance on various tasks (such as being helpful, harmless, and honest) depends heavily on the composition of their training data. However, selecting a data mixture that achieves strong performance across all tasks is challenging. Existing approaches rely on large ablation studies, heuristics, or human intuition, but these can be prohibitively expensive and suboptimal. We study this problem in the setting of preference optimization via DPO and introduce AutoMixAlign (AMA), a theoretically-grounded algorithm that adaptively mixes datasets during training to balance performance across tasks. AMA first trains \textit{specialist models} for each task to determine losses that correspond to strong task performance. Then, it trains a generalist model using a novel minimax optimization that prioritizes tasks for which generalist model losses deviate most from specialist model losses. To optimize this problem, we propose two algorithms: (1) AMA-R, which adaptively reweights the objective to prioritize tasks, and (2) AMA-S, which adaptively adjusts how much data is sampled from each task to prioritize tasks. Both algorithms achieve a convergence rate of $O(1/\sqrt{T})$ in the convex case. AMA-R's convergence result follows from Sagawa et al. (2019), and we provide a convergence proof for AMA-S using online learning techniques such as EXP3. We evaluate AMA on several multitask alignment setups and find that AMA outperforms the standard alignment approach -- which simply optimizes the total loss across all tasks -- and also outperforms model merging methods.

cs.LG↗

arXiv2025

ChatVLA-2: Vision-Language-Action Model with Open-World Embodied Reasoning from Pretrained Knowledge

Vision-language-action (VLA) models have emerged as the next generation of models in robotics. However, despite leveraging powerful pre-trained Vision-Language Models (VLMs), existing end-to-end VLA systems often lose key capabilities during fine-tuning as the model adapts to specific robotic tasks. We argue that a generalizable VLA model should retain and expand upon the VLM's core competencies: 1) Open-world embodied reasoning - the VLA should inherit the knowledge from VLM, i.e., recognize anything that the VLM can recognize, be capable of solving math problems, and possess visual-spatial intelligence, 2) Reasoning following - effectively translating the open-world reasoning into actionable steps for the robot. In this work, we introduce ChatVLA-2, a novel mixture-of-expert VLA model coupled with a specialized two-stage training pipeline designed to preserve the VLM's original strengths while enabling actionable reasoning. To validate our approach, we design a math-matching task wherein a robot interprets math problems written on a whiteboard and picks corresponding number cards from a table to solve equations. Remarkably, our method exhibits exceptional mathematical reasoning and OCR capabilities, despite these abilities not being explicitly trained within the VLA. Furthermore, we demonstrate that the VLA possesses strong spatial reasoning skills, enabling it to interpret novel directional instructions involving previously unseen objects. Overall, our method showcases reasoning and comprehension abilities that significantly surpass state-of-the-art imitation learning methods such as OpenVLA, DexVLA, and pi-zero. This work represents a substantial advancement toward developing truly generalizable robotic foundation models endowed with robust reasoning capacities.

cs.RO↗

arXiv2025

WorldEval: World Model as Real-World Robot Policies Evaluator

The field of robotics has made significant strides toward developing generalist robot manipulation policies. However, evaluating these policies in real-world scenarios remains time-consuming and challenging, particularly as the number of tasks scales and environmental conditions change. In this work, we demonstrate that world models can serve as a scalable, reproducible, and reliable proxy for real-world robot policy evaluation. A key challenge is generating accurate policy videos from world models that faithfully reflect the robot actions. We observe that directly inputting robot actions or using high-dimensional encoding methods often fails to generate action-following videos. To address this, we propose Policy2Vec, a simple yet effective approach to turn a video generation model into a world simulator that follows latent action to generate the robot video. We then introduce WorldEval, an automated pipeline designed to evaluate real-world robot policies entirely online. WorldEval effectively ranks various robot policies and individual checkpoints within a single policy, and functions as a safety detector to prevent dangerous actions by newly developed robot models. Through comprehensive paired evaluations of manipulation policies in real-world environments, we demonstrate a strong correlation between policy performance in WorldEval and real-world scenarios. Furthermore, our method significantly outperforms popular methods such as real-to-sim approach.

cs.RO↗

arXiv2025

MAPS: Motivation-Aware Personalized Search via LLM-Driven Consultation Alignment

Personalized product search aims to retrieve and rank items that match users' preferences and search intent. Despite their effectiveness, existing approaches typically assume that users' query fully captures their real motivation. However, our analysis of a real-world e-commerce platform reveals that users often engage in relevant consultations before searching, indicating they refine intents through consultations based on motivation and need. The implied motivation in consultations is a key enhancing factor for personalized search. This unexplored area comes with new challenges including aligning contextual motivations with concise queries, bridging the category-text gap, and filtering noise within sequence history. To address these, we propose a Motivation-Aware Personalized Search (MAPS) method. It embeds queries and consultations into a unified semantic space via LLMs, utilizes a Mixture of Attention Experts (MoAE) to prioritize critical semantics, and introduces dual alignment: (1) contrastive learning aligns consultations, reviews, and product features; (2) bidirectional attention integrates motivation-aware embeddings with user preferences. Extensive experiments on real and synthetic data show MAPS outperforms existing methods in both retrieval and ranking tasks.

cs.IR↗

arXiv2025

Efficient ANN-SNN Conversion with Error Compensation Learning

Artificial neural networks (ANNs) have demonstrated outstanding performance in numerous tasks, but deployment in resource-constrained environments remains a challenge due to their high computational and memory requirements. Spiking neural networks (SNNs) operate through discrete spike events and offer superior energy efficiency, providing a bio-inspired alternative. However, current ANN-to-SNN conversion often results in significant accuracy loss and increased inference time due to conversion errors such as clipping, quantization, and uneven activation. This paper proposes a novel ANN-to-SNN conversion framework based on error compensation learning. We introduce a learnable threshold clipping function, dual-threshold neurons, and an optimized membrane potential initialization strategy to mitigate the conversion error. Together, these techniques address the clipping error through adaptive thresholds, dynamically reduce the quantization error through dual-threshold neurons, and minimize the non-uniformity error by effectively managing the membrane potential. Experimental results on CIFAR-10, CIFAR-100, ImageNet datasets show that our method achieves high-precision and ultra-low latency among existing conversion methods. Using only two time steps, our method significantly reduces the inference time while maintains competitive accuracy of 94.75% on CIFAR-10 dataset under ResNet-18 structure. This research promotes the practical application of SNNs on low-power hardware, making efficient real-time processing possible.

cs.LG↗