arXiv subjects

Xi Zhao

Publications and source records attributed to Xi Zhao.

At least 37 records · Page 2Linked to original sources

arXiv2025

RAVR: Reference-Answer-guided Variational Reasoning for Large Language Models

Reinforcement learning (RL) can refine the reasoning abilities of large language models (LLMs), but critically depends on a key prerequisite: the LLM can already generate high-utility reasoning paths with non-negligible probability. For tasks beyond the LLM's current competence, such reasoning path can be hard to sample, and learning risks reinforcing familiar but suboptimal reasoning. We are motivated by the insight from cognitive science that Why is this the answer is often an easier question than What is the answer, as it avoids the heavy cognitive load of open-ended exploration, opting instead for explanatory reconstruction-systematically retracing the reasoning that links a question to its answer. We show that LLMs can similarly leverage answers to derive high-quality reasoning paths. We formalize this phenomenon and prove that conditioning on answer provably increases the expected utility of sampled reasoning paths, thereby transforming intractable problems into learnable ones. Building on this insight, we introduce RAVR (Reference-Answer-guided Variational Reasoning), an end-to-end framework that uses answer-conditioned reasoning as a variational surrogate for question-only reasoning. Experiments in both general and math domains demonstrate consistent improvements over strong baselines. We further analyze the reasoning behavior and find that RAVR reduces hesitation, strengthens conclusion consolidation, and promotes problem-specific strategies in reasoning.

cs.AI↗

arXiv2025

Learn while Unlearn: An Iterative Unlearning Framework for Generative Language Models

Recent advances in machine learning, particularly in Natural Language Processing (NLP), have produced powerful models trained on vast datasets. However, these models risk leaking sensitive information, raising privacy concerns. In response, regulatory measures such as the European Union's General Data Protection Regulation (GDPR) have driven increasing interest in Machine Unlearning techniques, which enable models to selectively forget specific data entries. Early unlearning approaches primarily relied on pre-processing methods, while more recent research has shifted towards training-based solutions. Despite their effectiveness, a key limitation persists: most methods require access to original training data, which is often unavailable. Additionally, directly applying unlearning techniques bears the cost of undermining the model's expressive capabilities. To address these challenges, we introduce the Iterative Contrastive Unlearning (ICU) framework, which consists of three core components: A Knowledge Unlearning Induction module designed to target specific knowledge for removal using an unlearning loss; A Contrastive Learning Enhancement module to preserve the model's expressive capabilities against the pure unlearning goal; And an Iterative Unlearning Refinement module that dynamically adjusts the unlearning process through ongoing evaluation and updates. Experimental results demonstrate the efficacy of our ICU method in unlearning sensitive information while maintaining the model's overall performance, offering a promising solution for privacy-conscious machine learning applications.

cs.LG↗

arXiv2025

MobiAgent: A Systematic Framework for Customizable Mobile Agents

With the rapid advancement of Vision-Language Models (VLMs), GUI-based mobile agents have emerged as a key development direction for intelligent mobile systems. However, existing agent models continue to face significant challenges in real-world task execution, particularly in terms of accuracy and efficiency. To address these limitations, we propose MobiAgent, a comprehensive mobile agent system comprising three core components: the MobiMind-series agent models, the AgentRR acceleration framework, and the MobiFlow benchmarking suite. Furthermore, recognizing that the capabilities of current mobile agents are still limited by the availability of high-quality data, we have developed an AI-assisted agile data collection pipeline that significantly reduces the cost of manual annotation. Compared to both general-purpose LLMs and specialized GUI agent models, MobiAgent achieves state-of-the-art performance in real-world mobile scenarios.

cs.MA↗

arXiv2025

Bridging BCI and Communications: A MIMO Framework for EEG-to-ECoG Wireless Channel Modeling

As a method to connect human brain and external devices, Brain-computer interfaces (BCIs) are receiving extensive research attention. Recently, the integration of communication theory with BCI has emerged as a popular trend, offering potential to enhance system performance and shape next-generation communications. A key challenge in this field is modeling the brain wireless communication channel between intracranial electrocorticography (ECoG) emitting neurons and extracranial electroencephalography (EEG) receiving electrodes. However, the complex physiology of brain challenges the application of traditional channel modeling methods, leaving relevant research in its infancy. To address this gap, we propose a frequency-division multiple-input multiple-output (MIMO) estimation framework leveraging simultaneous macaque EEG and ECoG recordings, while employing neurophysiology-informed regularization to suppress noise interference. This approach reveals profound similarities between neural signal propagation and multi-antenna communication systems. Experimental results show improved estimation accuracy over conventional methods while highlighting a trade-off between frequency resolution and temporal stability determined by signal duration. This work establish a conceptual bridge between neural interfacing and communication theory, accelerating synergistic developments in both fields.

eess.SP↗

arXiv2025

PBR3DGen: A VLM-guided Mesh Generation with High-quality PBR Texture

Generating high-quality physically based rendering (PBR) materials is important to achieve realistic rendering in the downstream tasks, yet it remains challenging due to the intertwined effects of materials and lighting. While existing methods have made breakthroughs by incorporating material decomposition in the 3D generation pipeline, they tend to bake highlights into albedo and ignore spatially varying properties of metallicity and roughness. In this work, we present PBR3DGen, a two-stage mesh generation method with high-quality PBR materials that integrates the novel multi-view PBR material estimation model and a 3D PBR mesh reconstruction model. Specifically, PBR3DGen leverages vision language models (VLM) to guide multi-view diffusion, precisely capturing the spatial distribution and inherent attributes of reflective-metalness material. Additionally, we incorporate view-dependent illumination-aware conditions as pixel-aware priors to enhance spatially varying material properties. Furthermore, our reconstruction model reconstructs high-quality mesh with PBR materials. Experimental results demonstrate that PBR3DGen significantly outperforms existing methods, achieving new state-of-the-art results for PBR estimation and mesh generation. More results and visualization can be found on our project page: https://pbr3dgen1218.github.io/.

cs.CV↗

arXiv2024

Exploring the Meaningfulness of Nearest Neighbor Search in High-Dimensional Space

Dense high dimensional vectors are becoming increasingly vital in fields such as computer vision, machine learning, and large language models (LLMs), serving as standard representations for multimodal data. Now the dimensionality of these vector can exceed several thousands easily. Despite the nearest neighbor search (NNS) over these dense high dimensional vectors have been widely used for retrieval augmented generation (RAG) and many other applications, the effectiveness of NNS in such a high-dimensional space remains uncertain, given the possible challenge caused by the "curse of dimensionality." To address above question, in this paper, we conduct extensive NNS studies with different distance functions, such as $L_1$ distance, $L_2$ distance and angular-distance, across diverse embedding datasets, of varied types, dimensionality and modality. Our aim is to investigate factors influencing the meaningfulness of NNS. Our experiments reveal that high-dimensional text embeddings exhibit increased resilience as dimensionality rises to higher levels when compared to random vectors. This resilience suggests that text embeddings are less affected to the "curse of dimensionality," resulting in more meaningful NNS outcomes for practical use. Additionally, the choice of distance function has minimal impact on the relevance of NNS. Our study shows the effectiveness of the embedding-based data representation method and can offer opportunity for further optimization of dense vector-related applications.

cs.LG↗

arXiv2024

Diverse 3D Human Pose Generation in Scenes based on Decoupled Structure

This paper presents a novel method for generating diverse 3D human poses in scenes with semantic control. Existing methods heavily rely on the human-scene interaction dataset, resulting in a limited diversity of the generated human poses. To overcome this challenge, we propose to decouple the pose and interaction generation process. Our approach consists of three stages: pose generation, contact generation, and putting human into the scene. We train a pose generator on the human dataset to learn rich pose prior, and a contact generator on the human-scene interaction dataset to learn human-scene contact prior. Finally, the placing module puts the human body into the scene in a suitable and natural manner. The experimental results on the PROX dataset demonstrate that our method produces more physically plausible interactions and exhibits more diverse human poses. Furthermore, experiments on the MP3D-R dataset further validates the generalization ability of our method.

cs.CV↗

arXiv2024

CBNet: A Plug-and-Play Network for Segmentation-Based Scene Text Detection

Recently, segmentation-based methods are quite popular in scene text detection, which mainly contain two steps: text kernel segmentation and expansion. However, the segmentation process only considers each pixel independently, and the expansion process is difficult to achieve a favorable accuracy-speed trade-off. In this paper, we propose a Context-aware and Boundary-guided Network (CBN) to tackle these problems. In CBN, a basic text detector is firstly used to predict initial segmentation results. Then, we propose a context-aware module to enhance text kernel feature representations, which considers both global and local contexts. Finally, we introduce a boundary-guided module to expand enhanced text kernels adaptively with only the pixels on the contours, which not only obtains accurate text boundaries but also keeps high speed, especially on high-resolution output maps. In particular, with a lightweight backbone, the basic detector equipped with our proposed CBN achieves state-of-the-art results on several popular benchmarks, and our proposed CBN can be plugged into several segmentation-based methods. Code is available at https://github.com/XiiZhao/cbn.pytorch.

cs.CV↗

arXiv2024

Compress3D: a Compressed Latent Space for 3D Generation from a Single Image

3D generation has witnessed significant advancements, yet efficiently producing high-quality 3D assets from a single image remains challenging. In this paper, we present a triplane autoencoder, which encodes 3D models into a compact triplane latent space to effectively compress both the 3D geometry and texture information. Within the autoencoder framework, we introduce a 3D-aware cross-attention mechanism, which utilizes low-resolution latent representations to query features from a high-resolution 3D feature volume, thereby enhancing the representation capacity of the latent space. Subsequently, we train a diffusion model on this refined latent space. In contrast to solely relying on image embedding for 3D generation, our proposed method advocates for the simultaneous utilization of both image embedding and shape embedding as conditions. Specifically, the shape embedding is estimated via a diffusion prior model conditioned on the image embedding. Through comprehensive experiments, we demonstrate that our method outperforms state-of-the-art algorithms, achieving superior performance while requiring less training data and time. Our approach enables the generation of high-quality 3D assets in merely 7 seconds on a single A100 GPU.

cs.CV↗

arXiv2023

Reconstructing 3D Human Pose from RGB-D Data with Occlusions

We propose a new method to reconstruct the 3D human body from RGB-D images with occlusions. The foremost challenge is the incompleteness of the RGB-D data due to occlusions between the body and the environment, leading to implausible reconstructions that suffer from severe human-scene penetration. To reconstruct a semantically and physically plausible human body, we propose to reduce the solution space based on scene information and prior knowledge. Our key idea is to constrain the solution space of the human body by considering the occluded body parts and visible body parts separately: modeling all plausible poses where the occluded body parts do not penetrate the scene, and constraining the visible body parts using depth data. Specifically, the first component is realized by a neural network that estimates the candidate region named the "free zone", a region carved out of the open space within which it is safe to search for poses of the invisible body parts without concern for penetration. The second component constrains the visible body parts using the "truncated shadow volume" of the scanned body point cloud. Furthermore, we propose to use a volume matching strategy, which yields better performance than surface matching, to match the human body with the confined region. We conducted experiments on the PROX dataset, and the results demonstrate that our method produces more accurate and plausible results compared with other methods.

cs.CV↗

arXiv2022

Shape Completion with Points in the Shadow

Single-view point cloud completion aims to recover the full geometry of an object based on only limited observation, which is extremely hard due to the data sparsity and occlusion. The core challenge is to generate plausible geometries to fill the unobserved part of the object based on a partial scan, which is under-constrained and suffers from a huge solution space. Inspired by the classic shadow volume technique in computer graphics, we propose a new method to reduce the solution space effectively. Our method considers the camera a light source that casts rays toward the object. Such light rays build a reasonably constrained but sufficiently expressive basis for completion. The completion process is then formulated as a point displacement optimization problem. Points are initialized at the partial scan and then moved to their goal locations with two types of movements for each point: directional movements along the light rays and constrained local movement for shape refinement. We design neural networks to predict the ideal point movements to get the completion results. We demonstrate that our method is accurate, robust, and generalizable through exhaustive evaluation and comparison. Moreover, it outperforms state-of-the-art methods qualitatively and quantitatively on MVP datasets.

cs.CV↗

arXiv2022

A Learned Index for Exact Similarity Search in Metric Spaces

Indexing is an effective way to support efficient query processing in large databases. Recently the concept of learned index, which replaces or complements traditional index structures with machine learning models, has been actively explored to reduce storage and search costs. However, accurate and efficient similarity query processing in high-dimensional metric spaces remains to be an open challenge. In this paper, we propose a novel indexing approach called LIMS that uses data clustering, pivot-based data transformation techniques and learned indexes to support efficient similarity query processing in metric spaces. In LIMS, the underlying data is partitioned into clusters such that each cluster follows a relatively uniform data distribution. Data redistribution is achieved by utilizing a small number of pivots for each cluster. Similar data are mapped into compact regions and the mapped values are totally ordinal. Machine learning models are developed to approximate the position of each data record on disk. Efficient algorithms are designed for processing range queries and nearest neighbor queries based on LIMS, and for index maintenance with dynamic updates. Extensive experiments on real-world and synthetic datasets demonstrate the superiority of LIMS compared with traditional indexes and state-of-the-art learned indexes.

cs.DB↗

arXiv2022

DB-LSH: Locality-Sensitive Hashing with Query-based Dynamic Bucketing

Among many solutions to the high-dimensional approximate nearest neighbor (ANN) search problem, locality sensitive hashing (LSH) is known for its sub-linear query time and robust theoretical guarantee on query accuracy. Traditional LSH methods can generate a small number of candidates quickly from hash tables but suffer from large index sizes and hash boundary problems. Recent studies to address these issues often incur extra overhead to identify eligible candidates or remove false positives, making query time no longer sub-linear. To address this dilemma, in this paper we propose a novel LSH scheme called DB-LSH which supports efficient ANN search for large high-dimensional datasets. It organizes the projected spaces with multi-dimensional indexes rather than using fixed-width hash buckets. Our approach can significantly reduce the space cost as by avoiding the need to maintain many hash tables for different bucket sizes. During the query phase of DB-LSH, a small number of high-quality candidates can be generated efficiently by dynamically constructing query-based hypercubic buckets with the required widths through index-based window queries. For a dataset of $n$ $d$-dimensional points with approximation ratio $c$, our rigorous theoretical analysis shows that DB-LSH achieves a smaller query cost ${O(n^{ρ^*} d\log n)}$, where ${ρ^*}$ is bounded by ${1/c^α}$ while the bound is ${1/c}$ in the existing work. An extensive range of experiments on real-world data demonstrates the superiority of DB-LSH over state-of-the-art methods on both efficiency and accuracy.

cs.DB↗

arXiv2021

Quantized orbital-chasing liquid metal heterodimers directed by an integrated pilot-wave field

A millimetric bouncing droplet sustained on a vibrating bath becomes a moving wave source (particle) through periodically interacting with the local wave field it generates during the droplet-bath impact. By virtue of such particle-wave duality, the macroscopic hydrodynamic system imitates enigmatic behaviors of the quantum realm. Here we show that it is possible to create an integrated pilot-wave field to better prescribe the droplet trajectories, via amplified bath capillarity. This is demonstrated with a liquid metal droplet-bath system in which the local wave field generated by droplet bouncing is superposed by the global wave field induced by bath meniscus oscillation. The resulting dual pilot-wave configuration enables a class of directional chasing motions of two bound dissimilar droplets (heterodimers) in multilevel hydrodynamic traps (orbits), featuring two quantized regime parameters, namely the interdroplet binding level and the orbit level. We investigate the dynamics of the vibrating liquid metal bath, with its level-split ring wave field and its peculiar vortex field being highlighted. We also rationalize the exotic droplet motions by considering the interdroplet particle-wave interactions mediated by the integrated pilot-wave field. It is revealed that a temporal bouncing phase shift between the two droplets in the heterodimers, due to size mismatch, gives rise to their horizontal propulsion, while their spatial binding regime exclusively determines the collective chasing direction. It is further evidenced that the horizontal in-orbit chasing motion is directly related to vertical droplet bouncing. Our findings unveil the integrated pilot-wave field as a trail towards improved droplet guiding, thereby extending the hydrodynamic particle-wave analogy to optical systems and beyond.

physics.flu-dyn↗

arXiv2021

PM-LSH: a fast and accurate in-memory framework for high-dimensional approximate NN and closest pair search

Nearest neighbor (NN) search is inherently computationally expensive in high-dimensional spaces due to the curse of dimensionality. As a well-known solution, locality-sensitive hashing (LSH) is able to answer c-approximate NN (c-ANN) queries in sublinear time with constant probability. Existing LSH methods focus mainly on building hash bucket-based indexing such that the candidate points can be retrieved quickly. However, existing coarse-grained structures fail to offer accurate distance estimation for candidate points, which translates into additional computational overhead when having to examine unnecessary points. This in turn reduces the performance of query processing. In contrast, we propose a fast and accurate in-memory LSH framework, called PM-LSH, that aims to compute c-ANN queries on large-scale, high-dimensional datasets. First, we adopt a simple yet effective PM-tree to index the data points. Second, we develop a tunable confidence interval to achieve accurate distance estimation and guarantee high result quality. Third, we propose an efficient algorithm on top of the PM-tree to improve the performance of computing c-ANN queries. In addition, we extend PM-LSH to support closest pair (CP) search in high-dimensional spaces. We again adopt the PM-tree to organize the points in a lowdimensional space, and we propose a branch and bound algorithm together with a radius pruning technique to improve the performance of computing c-approximate closest pair (c-ACP) queries. Extensive experiments with real-world data offer evidence that PM-LSH is capable of outperforming existing proposals with respect to both efficiency and accuracy for both NN and CP search.

cs.DB↗

arXiv2021

Semantic Data Augmentation for End-to-End Mandarin Speech Recognition

End-to-end models have gradually become the preferred option for automatic speech recognition (ASR) applications. During the training of end-to-end ASR, data augmentation is a quite effective technique for regularizing the neural networks. This paper proposes a novel data augmentation technique based on semantic transposition of the transcriptions via syntax rules for end-to-end Mandarin ASR. Specifically, we first segment the transcriptions based on part-of-speech tags. Then transposition strategies, such as placing the object in front of the subject or swapping the subject and the object, are applied on the segmented sentences. Finally, the acoustic features corresponding to the transposed transcription are reassembled based on the audio-to-text forced-alignment produced by a pre-trained ASR system. The combination of original data and augmented one is used for training a new ASR system. The experiments are conducted on the Transformer[2] and Conformer[3] based ASR. The results show that the proposed method can give consistent performance gain to the system. Augmentation related issues, such as comparison of different strategies and ratios for data combination are also investigated.

eess.AS↗

arXiv2021

REPOSE: Distributed Top-k Trajectory Similarity Search with Local Reference Point Tries

Trajectory similarity computation is a fundamental component in a variety of real-world applications, such as ridesharing, road planning, and transportation optimization. Recent advances in mobile devices have enabled an unprecedented increase in the amount of available trajectory data such that efficient query processing can no longer be supported by a single machine. As a result, means of performing distributed in-memory trajectory similarity search are called for. However, existing distributed proposals suffer from either computing resource waste or are unable to support the range of similarity measures that are being used. We propose a distributed in-memory management framework called REPOSE for processing top-k trajectory similarity queries on Spark. We develop a reference point trie (RP-Trie) index to organize trajectory data for local search. In addition, we design a novel heterogeneous global partitioning strategy to eliminate load imbalance in distributed settings. We report on extensive experiments with real-world data that offer insight into the performance of the solution, and show that the solution is capable of outperforming the state-of-the-art proposals.

cs.DB↗

arXiv2020

HeTDSE: GPU based program to solve full-dimensional-time-dependent Schrödinger equation for two-electron Helium subjected to strong laser fields

We present a GPU based openACC fortran program named HeTDSE, which provides an efficient way to investigate the non-perturbative electronic dynamics of helium subjected to a strong laser pulse by solving full-dimensional two-electron time dependent Schrödinger equation (TDSE). OpenACC is a directive-based programming model for accelerators without modifying the underlying CPU code itself. HeTDSE uses B-Spline basis sets expansion method to construct the radial part of the wavefunction, and the spherical harmonic functions is used to express for the angular part. Adams algorithm is employed for the time propagation. Our example shows HeTDSE running on an NVIDIA Kepler K20 GPU can outperform the one on an Intel E5-2640 single CPU core by a factor of 147.

physics.comp-ph↗