arXiv subjects

Feng Xiao

Publications and source records attributed to Feng Xiao.

At least 37 records · Page 2Linked to original sources

arXiv2026

EvolvR: Self-Evolving Pairwise Reasoning for Story Evaluation to Enhance Generation

Although the effectiveness of Large Language Models (LLMs) as judges (LLM-as-a-judge) has been validated, their performance remains limited in open-ended tasks, particularly in story evaluation. Accurate story evaluation is crucial not only for assisting human quality judgment but also for providing key signals to guide story generation. However, existing methods face a dilemma: prompt engineering for closed-source models suffers from poor adaptability, while fine-tuning approaches for open-source models lack the rigorous reasoning capabilities essential for story evaluation. To address this, we propose the Self-Evolving Pairwise Reasoning (EvolvR) framework. Grounded in pairwise comparison, the framework first self-synthesizes score-aligned Chain-of-Thought (CoT) data via a multi-persona strategy. To ensure data quality, these raw CoTs undergo a self-filtering process, utilizing multi-agents to guarantee their logical rigor and robustness. Finally, the evaluator trained on the refined data is deployed as a reward model to guide the story generation task. Experimental results demonstrate that our framework achieves state-of-the-art (SOTA) performance on three evaluation benchmarks including StoryER, HANNA and OpenMEVA. Furthermore, when served as a reward model, it significantly enhances the quality of generated stories, thereby fully validating the superiority of our self-evolving approach.

cs.CL↗

arXiv2026

Textual semantics and machine learning methods for data product pricing

Reasonable pricing of data products enables data trading platforms to maximize revenue and foster the growth of the data trading market. The textual semantics of data products are vital for pricing and contain significant value that remains largely underexplored. Therefore, to investigate how textual features influence data product pricing, we employ five prevalent text representation techniques to encode the descriptive text of data products. And then, we employ six machine learning methods to predict data product prices, including linear regression, neural networks, decision trees, support vector machines, random forests, and XGBoost. Our empirical design consists of two tasks: a regression task that predicts the continuous price of data products, and a classification task that discretizes price into ordered categories. Furthermore, we conduct feature importance analysis by the mRMR feature selection method and SHAP-based interpretability techniques. Based on empirical data from the AWA Data Exchange, we find that for predicting continuous prices, Word2Vec text representations capturing semantic similarity yield superior performance. In contrast, for price-tier classification tasks, simpler representations that do not rely on semantic similarity, such as Bag-of-Words and TF-IDF, perform better. SHAP analysis reveals that semantic features related to healthcare and demographics tend to increase prices, whereas those associated with weather and environmental topics are linked to lower prices. This analytical framework significantly enhances the interpretability of pricing models.

cs.CE↗

arXiv2026

Bumper Drone: Elastic Morphology Design for Aerial Physical Interaction

Aerial robots are evolving from avoiding obstacles to exploiting the environmental contact interactions for navigation, exploration and manipulation. A key challenge in such aerial physical interactions lies in handling uncertain contact forces on unknown targets, which typically demand accurate sensing and active control. We present a drone platform with elastic horns that enables touch-and-go manoeuvres - a self-regulated, consecutive bumping motion that allows the drone to maintain proximity to a wall without relying on active obstacle avoidance. It leverages environmental interaction as a form of embodied control, where low-level stabilisation and near-obstacle navigation emerge from the passive dynamic responses of the drone-obstacle system that resembles a mass-spring-damper system. Experiments show that the elastic horn can absorb impact energy while maintaining vehicle stability, reducing pitch oscillations by 38% compared to the rigid horn configuration. The lower horn arrangement was found to reduce pitch oscillations by approximately 54%. In addition to intermittent contact, the platform equipped with elastic horns also demonstrates stable, sustained contact with static objects, relying on a standard attitude PID controller.

cs.RO↗

arXiv2026

Distinguishing True Influence from Hyperprolificity with Citation Distance

Accurately evaluating scholarly influence is essential for fair academic assessment, yet traditional bibliometric indicators - dominated by publication and citation counts - often favor hyperprolific authors over those with deeper, long-term impact. We propose the x-index, a novel citation-based metric that conceptualizes citation as a process of knowledge diffusion and incorporates citation distance to reflect the structural reach of scholarly work. By weighting citations according to the collaborative proximity between citing and cited authors, the x-index captures both the depth and breadth of influence within evolving academic networks. Empirical analyses show that the x-index significantly improves the rankings of Turing Award recipients while reducing those of hyperprolific authors, better aligning rankings with recognized academic merit. It also demonstrates superior discriminatory power among early-career researchers and reveals stronger sensitivity to institutional research quality. These results suggest that the x-index offers a more equitable and forward-looking alternative to existing metrics, with practical applications in talent identification, funding decisions, and academic recommendation systems.

cs.DL↗

arXiv2026

IRPM: Intergroup Relative Preference Modeling for Pointwise Generative Reward Models

Generative Reward Models (GRMs) have demonstrated strong performance in reward modeling, due to their interpretability and potential for refinement through reinforcement learning (RL). However, widely used pairwise GRMs create a computational bottleneck in reinforcement learning from human feedback (RLHF), when calibrating or aggregating preference signals over n candidates, often incurring O(n^2) pairwise judgments. To address this issue, we propose Intergroup Relative Preference Modeling (IRPM), an RL-based method that extends the Bradley--Terry preference-learning paradigm via intergroup comparisons to train pointwise GRMs from pairwise preference data. IRPM derives pointwise reward for each response by contrasting groups of chosen vs. rejected samples, enabling pointwise scores comparable across candidate sets and O(n) reward evaluation for a variable number of candidates during RL training, while preserving interpretability and scalability. Experiments show that IRPM achieves state-of-the-art performance among pointwise GRMs on RM-Bench, JudgeBench and RewardBench, and approaches the performance of leading pairwise GRMs. In addition, IRPM achieves substantial gains in post-training evaluations, demonstrating its effectiveness.

cs.LG↗

arXiv2025

B2N3D: Progressive Learning from Binary to N-ary Relationships for 3D Object Grounding

Localizing 3D objects using natural language is essential for robotic scene understanding. The descriptions often involve multiple spatial relationships to distinguish similar objects, making 3D-language alignment difficult. Current methods only model relationships for pairwise objects, ignoring the global perceptual significance of n-ary combinations in multi-modal relational understanding. To address this, we propose a novel progressive relational learning framework for 3D object grounding. We extend relational learning from binary to n-ary to identify visual relations that match the referential description globally. Given the absence of specific annotations for referred objects in the training data, we design a grouped supervision loss to facilitate n-ary relational learning. In the scene graph created with n-ary relationships, we use a multi-modal network with hybrid attention mechanisms to further localize the target within the n-ary combinations. Experiments and ablation studies on the ReferIt3D and ScanRefer benchmarks demonstrate that our method outperforms the state-of-the-art, and proves the advantages of the n-ary relational perception in 3D localization.

cs.CV↗

arXiv2025

Can Visual Input Be Compressed? A Visual Token Compression Benchmark for Large Multimodal Models

Large multimodal models (LMMs) often suffer from severe inference inefficiency due to the large number of visual tokens introduced by image encoders. While recent token compression methods, such as pruning and merging, have shown promise in reducing redundancy, their evaluation remains fragmented and inconsistent. In this work, we present UniPruneBench, a unified and extensible benchmark for visual token pruning in multimodal LLMs. UniPruneBench provides standardized protocols across six ability dimensions and ten datasets, covering ten representative compression algorithms and three families of LMMs (LLaVA-v1.5, Intern-VL3, and Qwen2.5-VL). Beyond task accuracy, it incorporates system-level metrics such as runtime and prefilling latency to provide a holistic view. Our experiments uncover several key findings: (1) random pruning is a surprisingly strong baseline, (2) no single method consistently outperforms others across scenarios, (3) pruning sensitivity varies significantly across tasks, with OCR being most vulnerable, and (4) pruning ratio is the dominant factor governing performance degradation. We believe UniPruneBench will serve as a reliable foundation for future research on efficient multimodal modeling.

cs.CV↗

arXiv2025

MV-Debate: Multi-view Agent Debate with Dynamic Reflection Gating for Multimodal Harmful Content Detection in Social Media

Social media has evolved into a complex multimodal environment where text, images, and other signals interact to shape nuanced meanings, often concealing harmful intent. Identifying such intent, whether sarcasm, hate speech, or misinformation, remains challenging due to cross-modal contradictions, rapid cultural shifts, and subtle pragmatic cues. To address these challenges, we propose MV-Debate, a multi-view agent debate framework with dynamic reflection gating for unified multimodal harmful content detection. MV-Debate assembles four complementary debate agents, a surface analyst, a deep reasoner, a modality contrast, and a social contextualist, to analyze content from diverse interpretive perspectives. Through iterative debate and reflection, the agents refine responses under a reflection-gain criterion, ensuring both accuracy and efficiency. Experiments on three benchmark datasets demonstrate that MV-Debate significantly outperforms strong single-model and existing multi-agent debate baselines. This work highlights the promise of multi-agent debate in advancing reliable social intent detection in safety-critical online contexts.

cs.AI↗

arXiv2025

LSVG: Language-Guided Scene Graphs with 2D-Assisted Multi-Modal Encoding for 3D Visual Grounding

3D visual grounding aims to localize the unique target described by natural languages in 3D scenes. The significant gap between 3D and language modalities makes it a notable challenge to distinguish multiple similar objects through the described spatial relationships. Current methods attempt to achieve cross-modal understanding in complex scenes via a target-centered learning mechanism, ignoring the modeling of referred objects. We propose a novel 3D visual grounding framework that constructs language-guided scene graphs with referred object discrimination to improve relational perception. The framework incorporates a dual-branch visual encoder that leverages pre-trained 2D semantics to enhance and supervise the multi-modal 3D encoding. Furthermore, we employ graph attention to promote relationship-oriented information fusion in cross-modal interaction. The learned object representations and scene graph structure enable effective alignment between 3D visual content and textual descriptions. Experimental results on popular benchmarks demonstrate our superior performance compared to state-of-the-art methods, especially in handling the challenges of multiple similar distractors.

cs.CV↗

arXiv2025

Cyc3D: Fine-grained Controllable 3D Generation via Cycle Consistency Regularization

Despite the remarkable progress of 3D generation, achieving controllability, i.e., ensuring consistency between generated 3D content and input conditions like edge and depth, remains a significant challenge. Existing methods often struggle to maintain accurate alignment, leading to noticeable discrepancies. To address this issue, we propose \name{}, a new framework that enhances controllable 3D generation by explicitly encouraging cyclic consistency between the second-order 3D content, generated based on extracted signals from the first-order generation, and its original input controls. Specifically, we employ an efficient feed-forward backbone that can generate a 3D object from an input condition and a text prompt. Given an initial viewpoint and a control signal, a novel view is rendered from the generated 3D content, from which the extracted condition is used to regenerate the 3D content. This re-generated output is then rendered back to the initial viewpoint, followed by another round of control signal extraction, forming a cyclic process with two consistency constraints. \emph{View consistency} ensures coherence between the two generated 3D objects, measured by semantic similarity to accommodate generative diversity. \emph{Condition consistency} aligns the final extracted signal with the original input control, preserving structural or geometric details throughout the process. Extensive experiments on popular benchmarks demonstrate that \name{} significantly improves controllability, especially for fine-grained details, outperforming existing methods across various conditions (e.g., +14.17\% PSNR for edge, +6.26\% PSNR for sketch).

cs.CV↗

arXiv2025

Cognitive Hierarchy in Day-to-day Network Flow Dynamics

When making route decisions, travelers may engage in a certain degree of reasoning about what the others will do in the upcoming day, rendering yesterday's shortest routes less attractive. This phenomenon was manifested in a recent virtual experiment that mimicked travelers' repeated daily trip-making process. Unfortunately, prevailing day-to-day traffic dynamical models failed to faithfully reproduce the collected flow evolution data therein. To this end, we propose a day-to-day traffic behavior modeling framework based on the Cognitive Hierarchy theory, in which travelers with different levels of strategic-reasoning capabilities form their own beliefs about lower-step travelers' capabilities when choosing their routes. Two widely-studied day-to-day models, the Network Tatonnement Process dynamic and the Logit dynamic, are extended into the framework and studied as examples. Calibration of the virtual experiment is performed using the extended Network Tatonnement Process dynamic, which fits the experimental data reasonably well. We show that the two extended dynamics have multiple equilibria, one of which is the classical user equilibrium. While analyzing global stability is intractable due to the presence of multiple equilibria, local stabilities near equilibria are developed analytically and verified by numerical experiments. General insights on how key parameters affect the stability of user equilibria are unveiled.

econ.GN↗

arXiv2025

Domain-Specific Pruning of Large Mixture-of-Experts Models with Few-shot Demonstrations

Mixture-of-Experts (MoE) models achieve a favorable trade-off between performance and inference efficiency by activating only a subset of experts. However, the memory overhead of storing all experts remains a major limitation, especially in large-scale MoE models such as DeepSeek-R1(671B). In this study, we investigate domain specialization and expert redundancy in large-scale MoE models and uncover a consistent behavior we term few-shot expert localization, with only a few in-domain demonstrations, the model consistently activates a sparse and stable subset of experts on tasks within the same domain. Building on this observation, we propose a simple yet effective pruning framework, EASY-EP, that leverages a few domain-specific demonstrations to identify and retain only the most relevant experts. EASY-EP comprises two key components: output-aware expert importance assessment and expert-level token contribution estimation. The former evaluates the importance of each expert for the current token by considering the gating scores and L2 norm of the outputs of activated experts, while the latter assesses the contribution of tokens based on representation similarities before and after routed experts. Experiments on DeepSeek-R1 and DeepSeek-V3-0324 show that our method can achieve comparable performances and $2.99\times$ throughput under the same memory budget with full model with only half the experts.

cs.CL↗

arXiv2025

Fairness-aware Anomaly Detection via Fair Projection

Unsupervised anomaly detection is a critical task in many high-social-impact applications such as finance, healthcare, social media, and cybersecurity, where demographics involving age, gender, race, disease, etc, are used frequently. In these scenarios, possible bias from anomaly detection systems can lead to unfair treatment for different groups and even exacerbate social bias. In this work, first, we thoroughly analyze the feasibility and necessary assumptions for ensuring group fairness in unsupervised anomaly detection. Second, we propose a novel fairness-aware anomaly detection method FairAD. From the normal training data, FairAD learns a projection to map data of different demographic groups to a common target distribution that is simple and compact, and hence provides a reliable base to estimate the density of the data. The density can be directly used to identify anomalies while the common target distribution ensures fairness between different groups. Furthermore, we propose a threshold-free fairness metric that provides a global view for model's fairness, eliminating dependence on manual threshold selection. Experiments on real-world benchmarks demonstrate that our method achieves an improved trade-off between detection accuracy and fairness under both balanced and skewed data across different groups.

cs.LG↗

arXiv2025

TiSpell: A Semi-Masked Methodology for Tibetan Spelling Correction covering Multi-Level Error with Data Augmentation

Multi-level Tibetan spelling correction addresses errors at both the character and syllable levels within a unified model. Existing methods focus mainly on single-level correction and lack effective integration of both levels. Moreover, there are no open-source datasets or augmentation methods tailored for this task in Tibetan. To tackle this, we propose a data augmentation approach using unlabeled text to generate multi-level corruptions, and introduce TiSpell, a semi-masked model capable of correcting both character- and syllable-level errors. Although syllable-level correction is more challenging due to its reliance on global context, our semi-masked strategy simplifies this process. We synthesize nine types of corruptions on clean sentences to create a robust training set. Experiments on both simulated and real-world data demonstrate that TiSpell, trained on our dataset, outperforms baseline models and matches the performance of state-of-the-art approaches, confirming its effectiveness.

cs.CL↗

arXiv2025

GSsplat: Generalizable Semantic Gaussian Splatting for Novel-view Synthesis in 3D Scenes

The semantic synthesis of unseen scenes from multiple viewpoints is crucial for research in 3D scene understanding. Current methods are capable of rendering novel-view images and semantic maps by reconstructing generalizable Neural Radiance Fields. However, they often suffer from limitations in speed and segmentation performance. We propose a generalizable semantic Gaussian Splatting method (GSsplat) for efficient novel-view synthesis. Our model predicts the positions and attributes of scene-adaptive Gaussian distributions from once input, replacing the densification and pruning processes of traditional scene-specific Gaussian Splatting. In the multi-task framework, a hybrid network is designed to extract color and semantic information and predict Gaussian parameters. To augment the spatial perception of Gaussians for high-quality rendering, we put forward a novel offset learning module through group-based supervision and a point-level interaction module with spatial unit aggregation. When evaluated with varying numbers of multi-view inputs, GSsplat achieves state-of-the-art performance for semantic synthesis at the fastest speed.

cs.GR↗

arXiv2025

Global Rice Multi-Class Segmentation Dataset (RiceSEG): A Comprehensive and Diverse High-Resolution RGB-Annotated Images for the Development and Benchmarking of Rice Segmentation Algorithms

Developing computer vision-based rice phenotyping techniques is crucial for precision field management and accelerating breeding, thereby continuously advancing rice production. Among phenotyping tasks, distinguishing image components is a key prerequisite for characterizing plant growth and development at the organ scale, enabling deeper insights into eco-physiological processes. However, due to the fine structure of rice organs and complex illumination within the canopy, this task remains highly challenging, underscoring the need for a high-quality training dataset. Such datasets are scarce, both due to a lack of large, representative collections of rice field images and the time-intensive nature of annotation. To address this gap, we established the first comprehensive multi-class rice semantic segmentation dataset, RiceSEG. We gathered nearly 50,000 high-resolution, ground-based images from five major rice-growing countries (China, Japan, India, the Philippines, and Tanzania), encompassing over 6,000 genotypes across all growth stages. From these original images, 3,078 representative samples were selected and annotated with six classes (background, green vegetation, senescent vegetation, panicle, weeds, and duckweed) to form the RiceSEG dataset. Notably, the sub-dataset from China spans all major genotypes and rice-growing environments from the northeast to the south. Both state-of-the-art convolutional neural networks and transformer-based semantic segmentation models were used as baselines. While these models perform reasonably well in segmenting background and green vegetation, they face difficulties during the reproductive stage, when canopy structures are more complex and multiple classes are involved. These findings highlight the importance of our dataset for developing specialized segmentation models for rice and other crops.

eess.IV↗

arXiv2025

A low-dissipation reconstruction scheme for compressible single- and multi-phase flows based on artificial neural networks

Solving compressible flows containing both smooth and discontinuous flow structures remains a significant challenge for finite volume methods. Godunov-type finite volume methods are commonly used for numerical simulations of compressible flows. One of the key factors in obtaining high-quality solutions is high-fidelity spatial reconstruction. In this work, we introduce a new paradigm for constructing high-resolution hybrid reconstruction schemes for compressible flows. This approach generates training data based on BVD schemes for supervised learning and employs ANN to create an indicator that pre-selects the most suitable reconstruction scheme for each cell, achieving the lowest global numerical dissipation. The numerical schemes under this paradigm are more computationally efficient than similar schemes within the BVD framework, as each cell only requires constructing a single interpolation function. Following this paradigm, a novel low-dissipation reconstruction scheme based on the MUSCL-THINC-BVD scheme, named the deepMTBVD scheme, is proposed for compressible single- and multi-phase flows. The performance of the proposed scheme has been extensively verified through benchmark tests of single- and multi-phase compressible flows, where discontinuous and vortical flow structures, like shock waves, contact discontinuities, and material interfaces, as well as vortices and shear instabilities of different scales, coexist simultaneously. Numerical results indicate that the new deepMTBVD scheme performs as well as the original MUSCL-THINC-BVD scheme for numerical simulations of compressible flows while reducing computational time by up to 40%.

physics.flu-dyn↗

arXiv2025

Positivity-preserving hybrid DG/FV method for compressible Euler equations with stiff source terms

In numerical simulations of complex fluid dynamical problems, unphysical negative density or pressure may appear, causing blow-up of the computation. With the aim of obtaining positivity-preserving solutions with multi-scale resolution for solving strongly compressible flows, including hypersonic flows and stiff detonation waves, we present a positivity-preserving hybrid discontinuous Galerkin/finite volume (DG/FV) method. The approach is based on a priori and a posteriori computational methodology. The a priori computation utilizes relaxed boundary variation diminishing (BVD) algorithm to find troubled cells where the DG operators are replaced by the FV operators. The FV operators then deploy a hyperbolic tangent function in the reconstruction procedure to prevent unphysical values appearing in the flux evaluation. The a posteriori computation detects unphysical negative values in a simplified version of multidimensional optimal order detection (MOOD) paradigm and in the worst case applies the first-order FV Godunov scheme to guarantee the positivity of the solution. The a priori computation produces fewer oscillatory solutions, so that the a posteriori computation triggers less first-order evaluation. A technique of utilizing the tangent of hyperbola for interface capturing (THINC) with anti-diffusion effect, which is also referred to as the technique of adaptive reconstruction in this work, is suggested to reduce the numerical dissipation of the shock-capturing scheme. The current approaches retain the capability of the DG scheme to resolve small scales and the capability of the FV scheme to capture sharp discontinuities at the subgrid level.

physics.comp-ph↗