Xinyu Wang

Jingtao Li, Xinyu Wang, Hengwei Zhao et al.

Anomaly segmentation in high spatial resolution (HSR) remote sensing imagery is aimed at segmenting anomaly patterns of the earth deviating from normal patterns, which plays an important role in various Earth vision applications. However, it is a challenging task due to the complex distribution and the irregular shapes of objects, and the lack of abnormal samples. To tackle these problems, an anomaly segmentation model based on pixel descriptors (ASD) is proposed for anomaly segmentation in HSR imagery. Specifically, deep one-class classification is introduced for anomaly segmentation in the feature space with discriminative pixel descriptors. The ASD model incorporates the data argument for generating virtual ab-normal samples, which can force the pixel descriptors to be compact for normal data and meanwhile to be diverse to avoid the model collapse problems when only positive samples participated in the training. In addition, the ASD introduced a multi-level and multi-scale feature extraction strategy for learning the low-level and semantic information to make the pixel descriptors feature-rich. The proposed ASD model was validated using four HSR datasets and compared with the recent state-of-the-art models, showing its potential value in Earth vision applications.

Pengjie Wang, Kaile Zhang, Xinyu Wang et al.

Oracle bone script, one of the earliest known forms of ancient Chinese writing, presents invaluable research materials for scholars studying the humanities and geography of the Shang Dynasty, dating back 3,000 years. The immense historical and cultural significance of these writings cannot be overstated. However, the passage of time has obscured much of their meaning, presenting a significant challenge in deciphering these ancient texts. With the advent of Artificial Intelligence (AI), employing AI to assist in deciphering Oracle Bone Characters (OBCs) has become a feasible option. Yet, progress in this area has been hindered by a lack of high-quality datasets. To address this issue, this paper details the creation of the HUST-OBC dataset. This dataset encompasses 77,064 images of 1,588 individual deciphered characters and 62,989 images of 9,411 undeciphered characters, with a total of 140,053 images, compiled from diverse sources. The hope is that this dataset could inspire and assist future research in deciphering those unknown OBCs. All the codes and datasets are available at https://github.com/Yuliang-Liu/Open-Oracle.

Tianyu Zhang, Xinyu Wang, Lu Li et al.

While diffusion models have revolutionized text-to-image generation with their ability to synthesize realistic and diverse scenes, they continue to struggle to generate consistent and legible text within images. This shortcoming is commonly attributed to the locality bias inherent in diffusion-based generation, which limits their ability to model long-range spatial dependencies. In this paper, we introduce $\textbf{STRICT}$, a benchmark designed to systematically stress-test the ability of diffusion models to render coherent and instruction-aligned text in images. Our benchmark evaluates models across multiple dimensions: (1) the maximum length of readable text that can be generated; (2) the correctness and legibility of the generated text, and (3) the ratio of not following instructions for generating text. We evaluate several state-of-the-art models, including proprietary and open-source variants, and reveal persistent limitations in long-range consistency and instruction-following capabilities. Our findings provide insights into architectural bottlenecks and motivate future research directions in multimodal generative modeling. We release our entire evaluation pipeline at https://github.com/tianyu-z/STRICT-Bench.

Guoliang Dong, Haoyu Wang, Jun Sun et al.

By training on text in various languages, large language models (LLMs) typically possess multilingual support and demonstrate remarkable capabilities in solving tasks described in different languages. However, LLMs can exhibit linguistic discrimination due to the uneven distribution of training data across languages. That is, LLMs are hard to keep the consistency of responses when faced with the same task but depicted in different languages. In this study, we first explore the consistency in the LLMs' outputs responding to queries in various languages from two aspects: safety and quality. We conduct this analysis with two datasets (AdvBench and NQ) based on four LLMs (Llama2-13b, Gemma-7b, GPT-3.5-turbo and Gemini-pro). The results show that LLMs exhibit stronger human alignment capabilities with queries in English, French, Russian, and Spanish (only 1.04\% of harmful queries successfully jailbreak on average) compared to queries in Bengali, Georgian, Nepali and Maithili (27.7\% of harmful queries jailbreak successfully on average). Moreover, for queries in English, Danish, Czech and Slovenian, LLMs tend to produce responses with a higher quality (with 0.1494 $F_1$ score on average) compared to the other languages. Upon these findings, we propose LDFighter, a similarity-based voting, to mitigate the linguistic discrimination in LLMs. LDFighter ensures consistent service for different language speakers. We evaluate LDFighter with both benign queries and harmful queries. The results show that LDFighter not only significantly reduces the jailbreak success rate but also improve the response quality on average, demonstrating its effectiveness.

Xinyu Wang, Muhammad Ibrahim, Haitian Wang et al.

Accurate geo-registration of LiDAR point clouds remains a significant challenge in urban environments where Global Navigation Satellite System (GNSS) signals are denied or degraded. Existing methods typically rely on real-time GNSS and Inertial Measurement Unit (IMU) data, which require pre-calibration and assume stable signals. However, this assumption often fails in dense cities, resulting in localization errors. To address this, we propose a structured geo-registration method that accurately aligns LiDAR point clouds with satellite images, enabling frame-wise geo-registration and city-scale 3D reconstruction without prior localization. Our method uses a pre-trained Point Transformer to segment road points, then extracts road skeletons and intersections from the point cloud and the satellite image. Global alignment is achieved through rigid transformation using corresponding intersection points, followed by local non-rigid refinement with radial basis function (RBF) interpolation. Elevation discrepancies are corrected using terrain data from the Shuttle Radar Topography Mission (SRTM). To evaluate geo-registration accuracy, we measure the absolute distances between the roads extracted from the two modalities. Our method is validated on the KITTI benchmark and a newly collected dataset of Perth, Western Australia. On KITTI, our method achieves a mean planimetric alignment error of 0.69m, representing 50% improvement over the raw KITTI data. On Perth dataset, it achieves a mean planimetric error of 2.17m from GNSS values extracted from Google Maps, corresponding to 57.4% improvement over rigid alignment. Elevation correlation improved by 30.5% (KITTI) and 55.8% (Perth). A demonstration video is available at: https://youtu.be/0wkACAB-O6E.

Xinyu Wang, Muhammad Ibrahim, Atif Mansoor et al.

Road information extraction from 3D point clouds is useful for urban planning and traffic management. Existing methods often rely on local features and the refraction angle of lasers from kerbs, which makes them sensitive to variable kerb designs and issues in high-density areas due to data homogeneity. We propose an approach for extracting road points and fitting centrelines using a top-down view of LiDAR based ground-collected point clouds. This prospective view reduces reliance on specific kerb design and results in better road extraction. We first perform statistical outlier removal and density-based clustering to reduce noise from 3D point cloud data. Next, we perform ground point filtering using a grid-based segmentation method that adapts to diverse road scenarios and terrain characteristics. The filtered points are then projected onto a 2D plane, and the road is extracted by a skeletonisation algorithm. The skeleton is back-projected onto the 3D point cloud with calculated normals, which guide a region growing algorithm to find nearby road points. The extracted road points are then smoothed with the Savitzky-Golay filter to produce the final centreline. Our initial approach without post-processing of road skeleton achieved 67% in IoU by testing on the Perth CBD dataset with different road types. Incorporating the post-processing of the road skeleton improved the extraction of road points around the smoothed skeleton. The refined approach achieved a higher IoU value of 73% and with 23% reduction in the processing time. Our approach offers a generalised and computationally efficient solution that combines 3D and 2D processing techniques, laying the groundwork for future road reconstruction and 3D-to-2D point cloud alignment.

Maojia Song, Renhang Liu, Xinyu Wang et al.

RAG (Retrieval-Augmented Generation) systems and web agents are increasingly evaluated on multi-hop deep search tasks, yet current practice suffers from two major limitations. First, most benchmarks leak the reasoning path in the question text, allowing models to follow surface cues rather than discover reasoning chains autonomously. Second, evaluation is typically reduced to a single pass rate, which collapses diverse behaviours into one score and obscures whether failures stem from inadequate search, poor knowledge use, or inappropriate refusal. To address these issues, we present WebDetective, a benchmark of hint-free multi-hop questions paired with a controlled Wikipedia sandbox that ensures full traceability of model actions, and a holistic evaluation framework that separates search sufficiency, knowledge utilisation, and refusal behaviour. Our evaluation of 25 state-of-the-art models reveals systematic weaknesses across all architectures: models struggle with knowledge utilisation despite having sufficient evidence and demonstrate near-absent appropriate refusal when evidence is lacking. These patterns expose a fundamental gap: today's systems excel at executing given reasoning paths but fail when required to discover them. We develop an agentic workflow, EvidenceLoop, that explicitly targets the challenges our benchmark identifies, incorporating verification loops and systematic evidence tracking that improve both search and synthesis capabilities. This baseline demonstrates that WebDetective's diagnostic framework can guide concrete architectural improvements, establishing our benchmark as a critical tool for developing genuinely autonomous reasoning systems rather than pattern-following agents.

Xinyu Wang, Zixia Jia, Yong Jiang et al.

This paper describes the system used in submission from SHANGHAITECH team to the IWPT 2021 Shared Task. Our system is a graph-based parser with the technique of Automated Concatenation of Embeddings (ACE). Because recent work found that better word representations can be obtained by concatenating different types of embeddings, we use ACE to automatically find the better concatenation of embeddings for the task of enhanced universal dependencies. According to official results averaged on 17 languages, our system ranks 2nd over 9 teams.

Xinyu Wang, Yong Jiang, Zhaohui Yan et al.

Knowledge distillation is a critical technique to transfer knowledge between models, typically from a large model (the teacher) to a more fine-grained one (the student). The objective function of knowledge distillation is typically the cross-entropy between the teacher and the student's output distributions. However, for structured prediction problems, the output space is exponential in size; therefore, the cross-entropy objective becomes intractable to compute and optimize directly. In this paper, we derive a factorized form of the knowledge distillation objective for structured prediction, which is tractable for many typical choices of the teacher and student models. In particular, we show the tractability and empirical effectiveness of structural knowledge distillation between sequence labeling and dependency parsing models under four different scenarios: 1) the teacher and student share the same factorization form of the output structure scoring function; 2) the student factorization produces more fine-grained substructures than the teacher factorization; 3) the teacher factorization produces more fine-grained substructures than the student factorization; 4) the factorization forms from the teacher and the student are incompatible.

Xinyu Wang, Yong Jiang, Nguyen Bach et al.

Recent work proposes a family of contextual embeddings that significantly improves the accuracy of sequence labelers over non-contextual embeddings. However, there is no definite conclusion on whether we can build better sequence labelers by combining different kinds of embeddings in various settings. In this paper, we conduct extensive experiments on 3 tasks over 18 datasets and 8 languages to study the accuracy of sequence labeling with various embedding concatenations and make three observations: (1) concatenating more embedding variants leads to better accuracy in rich-resource and cross-domain settings and some conditions of low-resource settings; (2) concatenating additional contextual sub-word embeddings with contextual character embeddings hurts the accuracy in extremely low-resource settings; (3) based on the conclusion of (1), concatenating additional similar contextual embeddings cannot lead to further improvements. We hope these conclusions can help people build stronger sequence labelers in various settings.

Xinyu Wang, Yong Jiang, Kewei Tu

This paper presents the system used in our submission to the \textit{IWPT 2020 Shared Task}. Our system is a graph-based parser with second-order inference. For the low-resource Tamil corpus, we specially mixed the training data of Tamil with other languages and significantly improved the performance of Tamil. Due to our misunderstanding of the submission requirements, we submitted graphs that are not connected, which makes our system only rank \textbf{6th} over 10 teams. However, after we fixed this problem, our system is 0.6 ELAS higher than the team that ranked \textbf{1st} in the official results.

Xinyu Wang, Yong Jiang, Nguyen Bach et al.

Multilingual sequence labeling is a task of predicting label sequences using a single unified model for multiple languages. Compared with relying on multiple monolingual models, using a multilingual model has the benefit of a smaller model size, easier in online serving, and generalizability to low-resource languages. However, current multilingual models still underperform individual monolingual models significantly due to model capacity limitations. In this paper, we propose to reduce the gap between monolingual models and the unified multilingual model by distilling the structural knowledge of several monolingual models (teachers) to the unified multilingual model (student). We propose two novel KD methods based on structure-level information: (1) approximately minimizes the distance between the student's and the teachers' structure level probability distributions, (2) aggregates the structure-level knowledge to local distributions and minimizes the distance between two local probability distributions. Our experiments on 4 multilingual tasks with 25 datasets show that our approaches outperform several strong baselines and have stronger zero-shot generalizability than both the baseline model and teacher models.

Yuanyuan Jin, Wei Zhang, Xiangnan He et al.

Herb recommendation plays a crucial role in the therapeutic process of Traditional Chinese Medicine(TCM), which aims to recommend a set of herbs to treat the symptoms of a patient. While several machine learning methods have been developed for herb recommendation, they are limited in modeling only the interactions between herbs and symptoms, and ignoring the intermediate process of syndrome induction. When performing TCM diagnostics, an experienced doctor typically induces syndromes from the patient's symptoms and then suggests herbs based on the induced syndromes. As such, we believe the induction of syndromes, an overall description of the symptoms, is important for herb recommendation and should be properly handled. However, due to the ambiguity and complexity of syndrome induction, most prescriptions lack the explicit ground truth of syndromes. In this paper, we propose a new method that takes the implicit syndrome induction process into account for herb recommendation. Given a set of symptoms to treat, we aim to generate an overall syndrome representation by effectively fusing the embeddings of all the symptoms in the set, to mimic how a doctor induces the syndromes. Towards symptom embedding learning, we additionally construct a symptom-symptom graph from the input prescriptions for capturing the relations between symptoms; we then build graph convolution networks(GCNs) on both symptom-symptom and symptom-herb graphs to learn symptom embedding. Similarly, we construct a herb-herb graph and build GCNs on both herb-herb and symptom-herb graphs to learn herb embedding, which is finally interacted with the syndrome representation to predict the scores of herbs. In this way, more comprehensive representations can be obtained. We conduct extensive experiments on a public TCM dataset, showing significant improvements over state-of-the-art herb recommendation methods.