arXiv subjects

Xin Zhou

Publications and source records attributed to Xin Zhou.

At least 145 records · Page 8Linked to original sources

arXiv2025

Less is More: DocString Compression in Code Generation

The widespread use of Large Language Models (LLMs) in software engineering has intensified the need for improved model and resource efficiency. In particular, for neural code generation, LLMs are used to translate function/method signature and DocString to executable code. DocStrings which capture user re quirements for the code and used as the prompt for LLMs, often contains redundant information. Recent advancements in prompt compression have shown promising results in Natural Language Processing (NLP), but their applicability to code generation remains uncertain. Our empirical study show that the state-of-the-art prompt compression methods achieve only about 10% reduction, as further reductions would cause significant performance degradation. In our study, we propose a novel compression method, ShortenDoc, dedicated to DocString compression for code generation. Our extensive experiments on six code generation datasets, five open-source LLMs (1B to 10B parameters), and one closed-source LLM GPT-4o confirm that ShortenDoc achieves 25-40% compression while preserving the quality of generated code, outperforming other baseline methods at similar compression levels. The benefit of this research is to improve efficiency and reduce the cost while maintaining the quality of the generated code, especially when calling third-party APIs, and is able to reduce the token processing cost by 25-40%.

cs.SE↗

arXiv2025

Investigating the Perception of Translational Shape-Changing Haptic Interfaces

Shape-changing haptic interfaces (SCHIs) are a promising and emerging field. However, compared to more established stimulus modalities, such as vibration, there is sparse literature on the perception of dynamic shapes. Furthermore, the influence of properties such as grasp types and displacement magnitude/direction has not been formally evaluated. This work attempts to initiate a formal perceptual evaluation of SCHIs via a psychophysical user study involving a 1-DOF translational shape-changing interface that can move its body with 1.25-micrometer resolution. Participants completed a Method of Constant Stimulus study while holding the device with three different grasps. Stimuli direction occurred both toward and away from the thumb, while the standard stimuli varied between small (0.48 mm) and large (6 mm). Our results indicate that translational SCHIs should maximize the translation magnitude rather than the number of fingers in contact. We also demonstrated how to apply our findings to real-world applications via a simple 'paddle game', where we compared conventional linear mapping with non-linear mapping derived from our perceptual experiment outcomes between the device position and its represented value. Results indicate that the non-linear mapping was more effective, with improved error distribution. We hope this work inspires further formal perceptual investigation into other SCHI morphologies.

cs.HC↗

arXiv2025

MoTime: A Dataset Suite for Multimodal Time Series Forecasting

While multimodal data sources are increasingly available from real-world forecasting, most existing research remains on unimodal time series. In this work, we present MoTime, a suite of multimodal time series forecasting datasets that pair temporal signals with external modalities such as text, metadata, and images. Covering diverse domains, MoTime supports structured evaluation of modality utility under two scenarios: 1) the common forecasting task, where varying-length history is available, and 2) cold-start forecasting, where no historical data is available. Experiments show that external modalities can improve forecasting performance in both scenarios, with particularly strong benefits for short series in some datasets, though the impact varies depending on data characteristics. By making datasets and findings publicly available, we aim to support more comprehensive and realistic benchmarks in future multimodal time series forecasting research.

cs.LG↗

arXiv2025

CODE-DITING: A Reasoning-Based Metric for Functional Alignment in Code Evaluation

Trustworthy evaluation methods for code snippets play a crucial role in neural code generation. Traditional methods, which either rely on reference solutions or require executable test cases, have inherent limitation in flexibility and scalability. The recent LLM-as-Judge methodology offers a promising alternative by directly evaluating functional consistency between the problem description and the generated code. To systematically understand the landscape of these LLM-as-Judge methods, we conduct a comprehensive empirical study across three diverse datasets. Our investigation reveals the pros and cons of two categories of LLM-as-Judge methods: the methods based on general foundation models can achieve good performance but require complex prompts and lack explainability, while the methods based on reasoning foundation models provide better explainability with simpler prompts but demand substantial computational resources due to their large parameter sizes. To address these limitations, we propose CODE-DITING, a novel code evaluation method that balances accuracy, efficiency and explainability. We develop a data distillation framework that effectively transfers reasoning capabilities from DeepSeek-R1671B to our CODE-DITING 1.5B and 7B models, significantly enhancing evaluation explainability and reducing the computational cost. With the majority vote strategy in the inference process, CODE-DITING 1.5B outperforms all models with the same magnitude of parameters and achieves performance which would normally exhibit in a model with 5 times of parameter scale. CODE-DITING 7B surpasses GPT-4o and DeepSeek-V3 671B, even though it only uses 1% of the parameter volume of these large models. Further experiments show that CODEDITING is robust to preference leakage and can serve as a promising alternative for code evaluation.

cs.SE↗

arXiv2025

Semi-pessimistic Reinforcement Learning

Offline reinforcement learning (RL) aims to learn an optimal policy from pre-collected data. However, it faces challenges of distributional shift, where the learned policy may encounter unseen scenarios not covered in the offline data. Additionally, numerous applications suffer from a scarcity of labeled reward data. Relying on labeled data alone often leads to a narrow state-action distribution, further amplifying the distributional shift, and resulting in suboptimal policy learning. To address these issues, we first recognize that the volume of unlabeled data is typically substantially larger than that of labeled data. We then propose a semi-pessimistic RL method to effectively leverage abundant unlabeled data. Our approach offers several advantages. It considerably simplifies the learning process, as it seeks a lower bound of the reward function, rather than that of the Q-function or state transition function. It is highly flexible, and can be integrated with a range of model-free and model-based RL algorithms. It enjoys the guaranteed improvement when utilizing vast unlabeled data, but requires much less restrictive conditions. We compare our method with a number of alternative solutions, both analytically and numerically, and demonstrate its clear competitiveness. We further illustrate with an application to adaptive deep brain stimulation for Parkinson's disease.

cs.LG↗

arXiv2025

Response Uncertainty and Probe Modeling: Two Sides of the Same Coin in LLM Interpretability?

Probing techniques have shown promise in revealing how LLMs encode human-interpretable concepts, particularly when applied to curated datasets. However, the factors governing a dataset's suitability for effective probe training are not well-understood. This study hypothesizes that probe performance on such datasets reflects characteristics of both the LLM's generated responses and its internal feature space. Through quantitative analysis of probe performance and LLM response uncertainty across a series of tasks, we find a strong correlation: improved probe performance consistently corresponds to a reduction in response uncertainty, and vice versa. Subsequently, we delve deeper into this correlation through the lens of feature importance analysis. Our findings indicate that high LLM response variance is associated with a larger set of important features, which poses a greater challenge for probe models and often results in diminished performance. Moreover, leveraging the insights from response uncertainty analysis, we are able to identify concrete examples where LLM representations align with human knowledge across diverse domains, offering additional evidence of interpretable reasoning in LLMs.

cs.AI↗

arXiv2025

Bayesian ensemble learning for predicting health outcomes of multipollutant mixtures

We introduce the SoftBart approach from Bayesian ensemble learning to estimate the relationship between multipollutant mixtures and health on chronic exposures in epidemiology research. This approach offers several key advantages over existing methods: (1) it is computationally efficient and well-suited for analyzing large datasets; (2) it is flexible in estimating various correlated nonlinear functions simultaneously; and (3) it accurately identifies active variables within highly correlated multipollutant mixtures. Through simulations, we demonstrate the method's superiority by comparing its accuracy in estimating and quantifying uncertainties for both main and interaction effects with the commonly used method, BKMR. Last, we apply the method to analyze a multipollutant dataset with 10,110 participates from the Nurses' Health Study.

q-bio.QM↗

arXiv2025

GiFT: Gibbs Fine-Tuning for Code Generation

Training Large Language Models (LLMs) with synthetic data is a prevalent practice in code generation. A key approach is self-training, where LLMs are iteratively trained on self-generated correct code snippets. In this case, the self-generated codes are drawn from a conditional distribution, conditioned on a specific seed description. However, the seed description is not the only valid representation that aligns with its intended meaning. With all valid descriptions and codes forming a joint space, codes drawn from the conditional distribution would lead to an underrepresentation of the full description-code space. As such, we propose Gibbs Fine-Tuning (GiFT), a novel self-training method inspired by Gibbs sampling. GiFT allows self-generated data to be drawn from the marginal distribution of the joint space, thereby mitigating the biases inherent in conditional sampling. We provide a theoretical analysis demonstrating the potential benefits of fine-tuning LLMs with code derived from the marginal distribution. Furthermore, we propose a perplexity-based code selection method to mitigate the imbalanced long-tail distribution of the self-generated codes. Empirical evaluation of two LLMs across four datasets demonstrates that GiFT achieves superior performance, particularly on more challenging benchmarks. Source code is available at https://github.com/Alex-HaochenLi/GiFT.

cs.LG↗

arXiv2025

Extending Large Vision-Language Model for Diverse Interactive Tasks in Autonomous Driving

The Large Visual-Language Models (LVLMs) have significantly advanced image understanding. Their comprehension and reasoning capabilities enable promising applications in autonomous driving scenarios. However, existing research typically focuses on front-view perspectives and partial objects within scenes, struggling to achieve comprehensive scene understanding. Meanwhile, existing LVLMs suffer from the lack of mapping relationship between 2D and 3D and insufficient integration of 3D object localization and instruction understanding. To tackle these limitations, we first introduce NuInteract, a large-scale dataset with over 1.5M multi-view image language pairs spanning dense scene captions and diverse interactive tasks. Furthermore, we propose DriveMonkey, a simple yet effective framework that seamlessly integrates LVLMs with a spatial processor using a series of learnable queries. The spatial processor, designed as a plug-and-play component, can be initialized with pre-trained 3D detectors to improve 3D perception. Our experiments show that DriveMonkey outperforms general LVLMs, especially achieving a 9.86% notable improvement on the 3D visual grounding task. The dataset and code will be released at https://github.com/zc-zhao/DriveMonkey.

cs.CV↗

arXiv2025

Overcoming Intrinsic Dispersion Locking for Achieving Spatio-Spectral Selectivity with Misaligned Bi-metagratings

Spatio-spectral selectivity, the capability to select a single mode with a specific wavevector (angle) and wavelength, is imperative for light emission and imaging. Continuous band dispersion of a conventional periodic structure, however, sets up an intrinsic locking between wavevectors and wavelengths of photonic modes, making it difficult to single out just one mode. Here, we show that the radiation asymmetry of a photonic mode can be explored to tailor the transmission/reflection properties of a photonic structure, based on Fano interferences between the mode and the background. In particular, we find that a photonic system supporting a band dispersion with certain angle-dependent radiation-directionality can exhibit Fano-like perfect reflection at a single frequency and a single incident angle, thus overcoming the dispersion locking and enabling the desired spatio-spectral selectivity. We present a phase diagram to guide designing angle-controlled radiation-directionality and experimentally demonstrate double narrow Fano-like reflection in angular (5°) and wavelength (14 nm) bandwidths, along with high-contrast spatio-spectral selective imaging, using a misaligned bilayer metagrating with tens-of-nanometer-scale thin spacer. Our scheme promises new opportunities in applications in directional thermal emission, nonlocal beam shaping, augmented reality, precision bilayer nanofabrication, and biological spectroscopy.

physics.optics↗

arXiv2025

Learning Item Representations Directly from Multimodal Features for Effective Recommendation

Conventional multimodal recommender systems predominantly leverage Bayesian Personalized Ranking (BPR) optimization to learn item representations by amalgamating item identity (ID) embeddings with multimodal features. Nevertheless, our empirical and theoretical findings unequivocally demonstrate a pronounced optimization gradient bias in favor of acquiring representations from multimodal features over item ID embeddings. As a consequence, item ID embeddings frequently exhibit suboptimal characteristics despite the convergence of multimodal feature parameters. Given the rich informational content inherent in multimodal features, in this paper, we propose a novel model (i.e., LIRDRec) that learns item representations directly from these features to augment recommendation performance. Recognizing that features derived from each modality may capture disparate yet correlated aspects of items, we propose a multimodal transformation mechanism, integrated with modality-specific encoders, to effectively fuse features from all modalities. Moreover, to differentiate the influence of diverse modality types, we devise a progressive weight copying fusion module within LIRDRec. This module incrementally learns the weight assigned to each modality in synthesizing the final user or item representations. Finally, we utilize the powerful visual understanding of Multimodal Large Language Models (MLLMs) to convert the item images into texts and extract semantics embeddings upon the texts via LLMs. Empirical evaluations conducted on five real-world datasets validate the superiority of our approach relative to competing baselines. It is worth noting the proposed model, equipped with embeddings extracted from MLLMs and LLMs, can further improve the recommendation accuracy of NDCG@20 by an average of 4.21% compared to the original embeddings.

cs.IR↗

arXiv2025

AI-powered virtual eye: perspective, challenges and opportunities

We envision the "virtual eye" as a next-generation, AI-powered platform that uses interconnected foundation models to simulate the eye's intricate structure and biological function across all scales. Advances in AI, imaging, and multiomics provide a fertile ground for constructing a universal, high-fidelity digital replica of the human eye. This perspective traces the evolution from early mechanistic and rule-based models to contemporary AI-driven approaches, integrating in a unified model with multimodal, multiscale, dynamic predictive capabilities and embedded feedback mechanisms. We propose a development roadmap emphasizing the roles of large-scale multimodal datasets, generative AI, foundation models, agent-based architectures, and interactive interfaces. Despite challenges in interpretability, ethics, data processing and evaluation, the virtual eye holds the potential to revolutionize personalized ophthalmic care and accelerate research into ocular health and disease.

q-bio.TO↗

arXiv2025

Kubrick: Multimodal Agent Collaborations for Synthetic Video Generation

Text-to-video generation has been dominated by diffusion-based or autoregressive models. These novel models provide plausible versatility, but are criticized for improper physical motion, shading and illumination, camera motion, and temporal consistency. The film industry relies on manually-edited Computer-Generated Imagery (CGI) using 3D modeling software. Human-directed 3D synthetic videos address these shortcomings, but require tight collaboration between movie makers and 3D rendering experts. We introduce an automatic synthetic video generation pipeline based on Vision Large Language Model (VLM) agent collaborations. Given a language description of a video, multiple VLM agents direct various processes of the generation pipeline. They cooperate to create Blender scripts which render a video following the given description. Augmented with Blender-based movie making knowledge, the Director agent decomposes the text-based video description into sub-processes. For each sub-process, the Programmer agent produces Python-based Blender scripts based on function composing and API calling. The Reviewer agent, with knowledge of video reviewing, character motion coordinates, and intermediate screenshots, provides feedback to the Programmer agent. The Programmer agent iteratively improves scripts to yield the best video outcome. Our generated videos show better quality than commercial video generation models in five metrics on video quality and instruction-following performance. Our framework outperforms other approaches in a user study on quality, consistency, and rationality.

cs.CV↗

arXiv2025

Gas Transfer Between the Inner 3-kpc Disk and the Galactic Central Molecular Zone

We uncovered a more tilted molecular gas structure with highly negative velocities located near the dust lane. Our observations also show that the approaching gas flows from the overshoot process are captured by the bar gravitational and then flows towards the Galactic central molecular zone (CMZ) through the bar channel. The recycling gas from the overshoot effect, in conjunction with freshly accreted gas from the inner 3-kpc disk, accumulates significantly near R_GC~1/2R_bar and R_GC~2/3R_bar regions by adopting a bar length of ~3.2--3.4kpc. Importantly, within these regions, there are frequent collisions and substantial angular momentum exchanges between gas flows with different trajectories. In this scenario, the DISSIPATION processes arising from interactions between colliding flows, together with the varying torques induced by the nonaxisymmetric bar, effectively transfer the angular momentum of viscous gas outward, thereby driving the molecular gas to settle into the CMZ within ~3 orbital periods. A long-term gas inflow with an average rate of >1.1Msun/yr, coupled with intense transient accretion events that exceed the average rate by several times due to the overshoot effect, significantly regulates the gas distribution, physical properties, and dynamical evolution of the CMZ. These findings provide robust observational evidence for elucidating the intricate dynamics of molecular gas flows towards the CMZ. Our results show that gas dynamics has a significant impact on the secular evolution of both the Milky Way and the extragalactic gas-rich galaxies.

astro-ph.GA↗

arXiv2025

LMDepth: Lightweight Mamba-based Monocular Depth Estimation for Real-World Deployment

Monocular depth estimation provides an additional depth dimension to RGB images, making it widely applicable in various fields such as virtual reality, autonomous driving and robotic navigation. However, existing depth estimation algorithms often struggle to effectively balance performance and computational efficiency, which poses challenges for deployment on resource-constrained devices. To address this, we propose LMDepth, a lightweight Mamba-based monocular depth estimation network, designed to reconstruct high-precision depth information while maintaining low computational overhead. Specifically, we propose a modified pyramid spatial pooling module that serves as a multi-scale feature aggregator and context extractor, ensuring global spatial information for accurate depth estimation. Moreover, we integrate multiple depth Mamba blocks into the decoder. Designed with linear computations, the Mamba Blocks enable LMDepth to efficiently decode depth information from global features, providing a lightweight alternative to Transformer-based architectures that depend on complex attention mechanisms. Extensive experiments on the NYUDv2 and KITTI datasets demonstrate the effectiveness of our proposed LMDepth. Compared to previous lightweight depth estimation methods, LMDepth achieves higher performance with fewer parameters and lower computational complexity (measured by GFLOPs). We further deploy LMDepth on an embedded platform with INT8 quantization, validating its practicality for real-world edge applications.

cs.CV↗

arXiv2025

Effective Length Extrapolation via Dimension-Wise Positional Embeddings Manipulation

Large Language Models (LLMs) often struggle to process and generate coherent context when the number of input tokens exceeds the pre-trained length. Recent advancements in long-context extension have significantly expanded the context window of LLMs but require expensive overhead to train the large-scale models with longer context. In this work, we propose Dimension-Wise Positional Embeddings Manipulation (DPE), a training-free framework to extrapolate the context window of LLMs by diving into RoPE's different hidden dimensions. Instead of manipulating all dimensions equally, DPE detects the effective length for every dimension and finds the key dimensions for context extension. We reuse the original position indices with their embeddings from the pre-trained model and manipulate the key dimensions' position indices to their most effective lengths. In this way, DPE adjusts the pre-trained models with minimal modifications while ensuring that each dimension reaches its optimal state for extrapolation. DPE significantly surpasses well-known baselines such as YaRN and Self-Extend. DPE enables Llama3-8k 8B to support context windows of 128k tokens without continual training and integrates seamlessly with Flash Attention 2. In addition to its impressive extrapolation capability, DPE also dramatically improves the models' performance within training length, such as Llama3.1 70B, by over 18 points on popular long-context benchmarks RULER. When compared with commercial models, Llama 3.1 70B with DPE even achieves better performance than GPT-4-128K.

cs.CL↗

arXiv2025

SOFARI-R: High-Dimensional Manifold-Based Inference for Latent Responses

Data reduction with uncertainty quantification plays a key role in various multi-task learning applications, where large numbers of responses and features are present. To this end, a general framework of high-dimensional manifold-based SOFAR inference (SOFARI) was introduced recently in Zheng, Zhou, Fan and Lv (2024) for interpretable multi-task learning inference focusing on the left factor vectors and singular values exploiting the latent singular value decomposition (SVD) structure. Yet, designing a valid inference procedure on the latent right factor vectors is not straightforward from that of the left ones and can be even more challenging due to asymmetry of left and right singular vectors in the response matrix. To tackle these issues, in this paper we suggest a new method of high-dimensional manifold-based SOFAR inference for latent responses (SOFARI-R), where two variants of SOFARI-R are introduced. The first variant deals with strongly orthogonal factors by coupling left singular vectors with the design matrix and then appropriately rescaling them to generate new Stiefel manifolds. The second variant handles the more general weakly orthogonal factors by employing the hard-thresholded SOFARI estimates and delicately incorporating approximation errors into the distribution. Both variants produce bias-corrected estimators for the latent right factor vectors that enjoy asymptotically normal distributions with justified asymptotic variance estimates. We demonstrate the effectiveness of the newly suggested method using extensive simulation studies and an economic application.

stat.ME↗

arXiv2025

Context-Awareness and Interpretability of Rare Occurrences for Discovery and Formalization of Critical Failure Modes

Vision systems are increasingly deployed in critical domains such as surveillance, law enforcement, and transportation. However, their vulnerabilities to rare or unforeseen scenarios pose significant safety risks. To address these challenges, we introduce Context-Awareness and Interpretability of Rare Occurrences (CAIRO), an ontology-based human-assistive discovery framework for failure cases (or CP - Critical Phenomena) detection and formalization. CAIRO by design incentivizes human-in-the-loop for testing and evaluation of criticality that arises from misdetections, adversarial attacks, and hallucinations in AI black-box models. Our robust analysis of object detection model(s) failures in automated driving systems (ADS) showcases scalable and interpretable ways of formalizing the observed gaps between camera perception and real-world contexts, resulting in test cases stored as explicit knowledge graphs (in OWL/XML format) amenable for sharing, downstream analysis, logical reasoning, and accountability.

cs.CV↗