Author name cluster

Hai Lin

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

9 papers

2 author rows

AAAI Conference 2026 Conference Paper

FUSE: Fine-Grained and Semantic-Aware Learning for Unified Image Understanding and Generation

Peng Zhang
Wanggui He
Mushui Liu
Wenyi Xiao
Siyu Zou
Yuan Li
Xingjian Wang
Guanghao Zhang

Recent unified models have demonstrated that the reasoning capacity of Multimodal Large Language Models (MLLMs) can be leveraged to facilitate diffusion-based image generation with impressive flexibility and performance. However, approaches that rely heavily on MLLMs for high-level semantic encoding often struggle with fine-grained visual tasks like image editing and virtual try-on. To address this gap, we propose FUSE, a unified framework excelling at both high-level vision–language understanding and fine-grained generation. First, we introduce a Semantic-to-Detail Connector that pre-aligns fine-grained visual features with the MLLM's semantic space. This design counteracts the low-level information loss inherent in MLLM encodings, creating a unified representation that steers the diffusion process with both global semantics and rich local details. Second, to further enhance semantic awareness and detail preservation, we introduce Adaptive-GRPO, a post-training objective that dynamically balances semantic coherence against pixel-level fidelity. The integration of these two innovations allows FUSE to generate images that are both semantically faithful and visually fine-grained. Comprehensive experiments on text-to-image and instruction-guided editing benchmarks show that FUSE significantly outperforms existing unified baselines, achieving 0.89 on Geneval, 0.65 on WISE, and 3.88 on ImageEdit.

PDF Details DOI

AAAI Conference 2025 Conference Paper

Frozen Language Models Are Gradient Coherence Rectifiers in Vision Transformers

Lichen Bai
Zixuan Xiong
Hai Lin
Guangwei Xu
Xiangjin Xie
Ruijie Guo
Zhanhui Kang
Hai-Tao Zheng

Large language models (LLMs) have demonstrated remarkable performance in multimodal tasks even with frozen LLM Block and only a few trainable parameters. However, the underlying mechanisms of how LLMs enhance multimodal performance remains unclear. In this work, we focus on the phenomenon that ``Merely concatenating a frozen LLM block to the Vision Transformer (ViT) encoder can yield significant performance enhancements. Moreover, the choice of LLM block and insertion position can have a substantial impact, leading to varying degrees of improvement''. We analyze the optimization of the training process from the perspective of gradient dynamics and find that frozen LLM blocks act as gradient coherence rectifiers, aligning the gradients of different samples more closely during training. Furthermore, we demonstrate that the representation similarity between the inserted LLM block and the adjacent ViT block influences performance, with greater similarity tending to yield larger positive gains. Through these findings, we can justify the selection of suitable LLM blocks to be inserted at appropriate positions, and introduce additional gradient backpropagation paths by incorporating LLM blocks, could improve the performance of vanilla ViT through the rectification effect of gradient consistency during the training process, without the need to add LLM blocks during inference. Our experiments demonstrate the effectiveness of this strategy, making the practical application of the gradient rectification effect feasible.

PDF Details DOI

ECAI Conference 2025 Conference Paper

LexSemBridge: Fine-Grained Dense Representation Enhancement Through Token-Aware Embedding Augmentation

Shaoxiong Zhan
Hai Lin
Hongming Tan
Xiaodong Cai
Hai-Tao Zheng
Xin Su
Zifei Shan
Ruitong Liu

As queries in retrieval-augmented generation (RAG) pipelines powered by large language models (LLMs) become increasingly complex and diverse, dense retrieval models have demonstrated strong performance in semantic matching. Nevertheless, they often struggle with fine-grained retrieval tasks, where precise keyword alignment and span-level localization are required, even in cases with high lexical overlap that would intuitively suggest easier retrieval. To systematically evaluate this limitation, we introduce two targeted tasks, keyword retrieval and part-of-passage retrieval, designed to simulate practical fine-grained scenarios. Motivated by these observations, we propose LexSemBridge, a unified framework that enhances dense query representations through fine-grained, input-aware vector modulation. LexSemBridge constructs latent enhancement vectors from input tokens using three paradigms: Statistical (SLR), Learned (LLR), and Contextual (CLR), and integrates them with dense embeddings via element-wise interaction. Theoretically, we show that this modulation preserves the semantic direction while selectively amplifying discriminative dimensions. LexSemBridge operates as a plug-in without modifying the backbone encoder and naturally extends to both text and vision modalities. Extensive experiments across semantic and fine-grained retrieval tasks validate the effectiveness and generality of our approach. All code and models are publicly available at https: //github. com/Jasaxion/LexSemBridge/

Details

AAAI Conference 2025 Conference Paper

SIGraph: Saliency Image-Graph Network for Retinal Disease Classification in Fundus Image

Peng Zhang
Yuan Li
Haotian Song
Yankai Jiang
Yubo Tao
Hai Lin
Hongguang Cui

An efficient and precise diagnosis of retinal diseases is a fundamental goal for auxiliary diagnostic systems in ophthalmology. Inspired by the importance of scattered subtle lesions in manual retinal disease diagnosis, recent research has achieved state-of-the-art performance by mining information related to subtle lesions, including their texture and shape. However, the spatial distribution patterns of subtle lesion areas, which are also crucial in manual diagnosis, have been overlooked in existing research. Neglecting these spatial distribution patterns (e.g., the ring distribution of microaneurysms in diabetic macular edema) may negatively impact the diagnostic process. In this paper, we introduce the Saliency-Image-Graph (SIGraph) network to capture the spatial distribution patterns of lesion areas. We first employ saliency-based perception to identify latent lesion pixels. Subsequently, we propose a novel image-graph block to efficiently capture the global distribution of abundant lesion pixels with minimal information loss. By leveraging additional distribution patterns, SIGraph achieves state-of-the-art performance with at least a 1.5% performance gain across three datasets. Furthermore, ablation studies demonstrate that our image-graph block can be integrated into other visual backbones and effectively boost performance.

PDF Details DOI

GandALF Workshop 2022 Workshop Paper

CryptoSolve: Towards a Tool for the Symbolic Analysis of Cryptographic Algorithms

Dalton Chichester
Wei Du
Raymond Kauffman
Hai Lin
Christopher Lynch
Andrew M. Marshall
Catherine A. Meadows
Paliath Narendran

Recently, interest has been emerging in the application of symbolic techniques to the specification and analysis of cryptosystems. These techniques, when accompanied by suitable proofs of soundness/completeness, can be used both to identify insecure cryptosystems and prove sound ones secure. But although a number of such symbolic algorithms have been developed and implemented, they remain scattered throughout the literature. In this paper, we present a tool, CryptoSolve, which provides a common basis for specification and implementation of these algorithms, CryptoSolve includes libraries that provide the term algebras used to express symbolic cryptographic systems, as well as implementations of useful algorithms, such as unification and variant generation. In its current initial iteration, it features several algorithms for the generation and analysis of cryptographic modes of operation, which allow one to use block ciphers to encrypt messages more than one block long. The goal of our work is to continue expanding the tool in order to consider additional cryptosystems and security questions, as well as extend the symbolic libraries to increase their applicability.

PDF Details DOI

IJCAI Conference 2022 Conference Paper

SatFormer: Saliency-Guided Abnormality-Aware Transformer for Retinal Disease Classification in Fundus Image

Yankai Jiang
Ke Xu
Xinyue Wang
Yuan Li
Hongguang Cui
Yubo Tao
Hai Lin

Automatic and accurate retinal disease diagnosis is critical to guide proper therapy and prevent potential vision loss. Previous works simply exploit the most discriminative features while ignoring the pathological visual clues of scattered subtle lesions. Therefore, without a comprehensive understanding of features from different lesion regions, they are vulnerable to noise from complex backgrounds and suffer from misclassification failures. In this paper, we address these limitations with a novel saliency-guided abnormality-aware transformer which explicitly captures the correlation between different lesion features from a global perspective with enhanced pathological semantics. The model has several merits. First, we propose a saliency enhancement module (SEM) which adaptively integrates disease related semantics and highlights potentially salient lesion regions. Second, to the best of our knowledge, this is the first work to explore comprehensive lesion feature dependencies via a tailored efficient self-attention. Third, with the saliency enhancement module and abnormality-aware attention, we propose a new variant of Vision Transformer models, called SatFormer, which outperforms the state-of-the-art methods on two public retinal disease classification benchmarks. Ablation study shows that the proposed components can be easily embedded into any Vision Transformers via a plug-and-play manner and effectively boost the performance.

PDF Details DOI

YNICL Journal 2020 Journal Article

Functional connectivity markers of depression in advanced Parkinson's disease

Hai Lin
Xiaodong Cai
Doudou Zhang
Jiali Liu
Peng Na
Weiping Li

BACKGROUND: Depression is a common comorbid condition in Parkinson's disease and a major contributor to poor quality of life. Despite this, depression in PD is under-diagnosed due to overlapping symptoms and difficulties in the assessment of depression in cognitively impaired old patients. OBJECTIVES: This study is to explore functional connectivity markers of depression in PD patients using resting-state fMRI and help diagnose whether patients have depression or not. METHODS: We reviewed 156 advanced PD patients (duration > 5 years; 59 depressed ones) and 45 healthy control subjects who underwent a resting-state fMRI scanning. Functional connectivity analysis was employed to characterize intrinsic connectivity networks using group independent component analysis and extract connectivity features. Features were put into an all-relevant feature selection procedure within cross-validation loops, to identify features with significant discriminative power for classification. Random forest classifiers were built for depression diagnosis, on the basis of identified features. RESULTS: 42 intrinsic connectivity networks were identified and arranged into subcortical, auditory, somatomotor, visual, cognitive control, default-mode and cerebellar networks. Six features were significantly relevant to classification. They were connectivity within posterior cingulate cortex, within insula, between posterior cingulate cortex and insula/hippocampus+amygdala, between insula and precuneus, and between superior parietal lobule and medial prefrontal cortex. The mean accuracy achieved with classifiers to discriminate depressed patients from the non-depressed was 82.4%. CONCLUSIONS: Our findings provide preliminary evidence that resting-state functional connectivity can characterize depressed PD patients and help distinguish them from non-depressed ones.

Details DOI

IROS Conference 2019 Conference Paper

Stability and Gait Switching of Underactuated Biped Walkers

Martin Fevre
Hai Lin
James P. Schmiedeler

This paper introduces a new gait switching approach for underactuated biped walkers. The switching condition relies on a reduced region of attraction — that of the unactuated dynamics — which is shown to be sufficient to predict falls. A gait transition is accordingly stable (as in “the robot does not fall if the biped’s state is within the reduced region of attraction of the switched-in gait when switching takes place. Two- and five-link biped models complete a sequence of random gait transitions using the switching logic. The condition is also used to enlarge the full region of stability of the five-link model by embedding feedback-stabilized trajectories from a gait library in a mapping. The mapping stitches stable trajectories to the orbit of the desired gait based on each gait’s reduced region of attraction. This improves the robustness of the five-link model walking on uneven terrain without ground perception regardless of the size of the gait library.

Details

EAAI Journal 2017 Journal Article

Spatially–temporally online fault detection using timed multivariate statistical logic

Kangling Liu
Hai Lin
Zhengshun Fei
Jun Liang

This work develops an extension of temporal logic called timed multivariate statistical logic (TMSL) that can specify not only spatial features but also temporal dynamics of systems in a formal way. A purely data-based algorithm is presented to automatically learn the TMSL from process data. First, the principal component analysis (PCA) method is used to extract spatial features among all available process data. Next, based on these spatial features, a large margin fuzzy c-means method is developed to automatically discover a set of meaningful regions called Regions-of-Interest. As a result, the data space is partitioned into a set of Regions-of-Interest. Then, a temporally-annotated automaton for TMSL is generated with these discovered Regions-of-Interest. Finally, a PCA-based spatial monitor and a TMSL-based temporal monitor are further developed for online fault detection. For performance validation, the proposed online fault detection method is demonstrated in three application studies.

Details DOI