Tian Liu

Luoming Zhang, Yuwei Ren, Kui Zhang et al.

Matrix inversion in chunk-wise parallel linear attention is a major bottleneck for long-context modeling, particularly on NPUs, where forward-substitution-based methods exhibit limited parallelism and poor hardware utilization. We propose a fast, Matrix Multiplication (MatMul)-based algorithm tailored for strictly lower-triangular matrices arising in chunk-wise linear attention. Motivated by the rapid growth of Neumann-series terms and the diagonal concentration of the inverse matrix, we employ a truncated Neumann expansion with structural masking and parallel residual correction to eliminate sequential dependencies. We further extend our method to low-bits INT by mitigating the dynamic range expansion arising from repeated matrix power operations, and adapt the approximation order and residual step to the chunk size to minimize computational cost while preserving the model's accuracy. Experiments on Qwen3.5-family models demonstrate up to 5$\times$ kernel-level speedup and a 20% reduction in decode-layer overhead, while preserving accuracy under both floating-point and low-precision inference. Our method offers an efficient and hardware-friendly solution for scalable linear attention.

Mingzhe Hu, Jiahan Zhang, Luke Matkovic et al.

Motivation: Medical image analysis involves tasks to assist physicians in qualitative and quantitative analysis of lesions or anatomical structures, significantly improving the accuracy and reliability of diagnosis and prognosis. Traditionally, these tasks are finished by physicians or medical physicists and lead to two major problems: (i) low efficiency; (ii) biased by personal experience. In the past decade, many machine learning methods have been applied to accelerate and automate the image analysis process. Compared to the enormous deployments of supervised and unsupervised learning models, attempts to use reinforcement learning in medical image analysis are scarce. This review article could serve as the stepping-stone for related research. Significance: From our observation, though reinforcement learning has gradually gained momentum in recent years, many researchers in the medical analysis field find it hard to understand and deploy in clinics. One cause is lacking well-organized review articles targeting readers lacking professional computer science backgrounds. Rather than providing a comprehensive list of all reinforcement learning models in medical image analysis, this paper may help the readers to learn how to formulate and solve their medical image analysis research as reinforcement learning problems. Approach & Results: We selected published articles from Google Scholar and PubMed. Considering the scarcity of related articles, we also included some outstanding newest preprints. The papers are carefully reviewed and categorized according to the type of image analysis task. We first review the basic concepts and popular models of reinforcement learning. Then we explore the applications of reinforcement learning models in landmark detection. Finally, we conclude the article by discussing the reviewed reinforcement learning approaches' limitations and possible improvements.

Tian Liu, Liuyi Jin, Radu Stoleru et al.

Current state-of-the-art residential irrigation systems, such as WaterMyYard, rely on rainfall data from nearby weather stations to adjust irrigation amounts. However, the accuracy of rainfall data is compromised by the limited spatial resolution of rain gauges and the significant variability of hyperlocal rainfall, leading to substantial water waste. To improve irrigation efficiency, we developed a cost-effective irrigation system, dubbed ERIC, which employs machine learning models to estimate rainfall from commodity doorbell camera footage and optimizes irrigation schedules without human intervention. Specifically, we: a) designed novel visual and audio features with lightweight neural network models to infer rainfall from the camera at the edge, preserving user privacy; b) built a complete end-to-end irrigation system on Raspberry Pi 4, costing only \$75. We deployed the system across five locations (collecting over 750 hours of video) with varying backgrounds and light conditions. Comprehensive evaluation validates that ERIC achieves state-of-the-art rainfall estimation performance ($\sim$ 5mm/day), saving 9,112 gallons/month of water, translating to \$28.56/month in utility savings. Data and code are available at https://github.com/LENSS/ERIC-BuildSys2024.git

Shaoyan Pan, Chih-Wei Chang, Junbo Peng et al.

This study aims to develop a novel Cycle-guided Denoising Diffusion Probability Model (CG-DDPM) for cross-modality MRI synthesis. The CG-DDPM deploys two DDPMs that condition each other to generate synthetic images from two different MRI pulse sequences. The two DDPMs exchange random latent noise in the reverse processes, which helps to regularize both DDPMs and generate matching images in two modalities. This improves image-to-image translation ac-curacy. We evaluated the CG-DDPM quantitatively using mean absolute error (MAE), multi-scale structural similarity index measure (MSSIM), and peak sig-nal-to-noise ratio (PSNR), as well as the network synthesis consistency, on the BraTS2020 dataset. Our proposed method showed high accuracy and reliable consistency for MRI synthesis. In addition, we compared the CG-DDPM with several other state-of-the-art networks and demonstrated statistically significant improvements in the image quality of synthetic MRIs. The proposed method enhances the capability of current multimodal MRI synthesis approaches, which could contribute to more accurate diagnosis and better treatment planning for patients by synthesizing additional MRI modalities.

Huiqiao Xie, Yang Lei, Yabo Fu et al.

CBCTs in image-guided radiotherapy provide crucial anatomy information for patient setup and plan evaluation. Longitudinal CBCT image registration could quantify the inter-fractional anatomic changes. The purpose of this study is to propose an unsupervised deep learning based CBCT-CBCT deformable image registration. The proposed deformable registration workflow consists of training and inference stages that share the same feed-forward path through a spatial transformation-based network (STN). The STN consists of a global generative adversarial network (GlobalGAN) and a local GAN (LocalGAN) to predict the coarse- and fine-scale motions, respectively. The network was trained by minimizing the image similarity loss and the deformable vector field (DVF) regularization loss without the supervision of ground truth DVFs. During the inference stage, patches of local DVF were predicted by the trained LocalGAN and fused to form a whole-image DVF. The local whole-image DVF was subsequently combined with the GlobalGAN generated DVF to obtain final DVF. The proposed method was evaluated using 100 fractional CBCTs from 20 abdominal cancer patients in the experiments and 105 fractional CBCTs from a cohort of 21 different abdominal cancer patients in a holdout test. Qualitatively, the registration results show great alignment between the deformed CBCT images and the target CBCT image. Quantitatively, the average target registration error (TRE) calculated on the fiducial markers and manually identified landmarks was 1.91+-1.11 mm. The average mean absolute error (MAE), normalized cross correlation (NCC) between the deformed CBCT and target CBCT were 33.42+-7.48 HU, 0.94+-0.04, respectively. This promising registration method could provide fast and accurate longitudinal CBCT alignment to facilitate inter-fractional anatomic changes analysis and prediction.

Yupei Zhang, Xianjin Dai, Zhen Tian et al.

This study proposed a deep learning-based tracking method for ultrasound (US) image-guided radiation therapy. The proposed cascade deep learning model is composed of an attention network, a mask region-based convolutional neural network (mask R-CNN), and a long short-term memory (LSTM) network. The attention network learns a mapping from a US image to a suspected area of landmark motion in order to reduce the search region. The mask R-CNN then produces multiple region-of-interest (ROI) proposals in the reduced region and identifies the proposed landmark via three network heads: bounding box regression, proposal classification, and landmark segmentation. The LSTM network models the temporal relationship among the successive image frames for bounding box regression and proposal classification. To consolidate the final proposal, a selection method is designed according to the similarities between sequential frames. The proposed method was tested on the liver US tracking datasets used in the Medical Image Computing and Computer Assisted Interventions (MICCAI) 2015 challenges, where the landmarks were annotated by three experienced observers to obtain their mean positions. Five-fold cross-validation on the 24 given US sequences with ground truths shows that the mean tracking error for all landmarks is 0.65+/-0.56 mm, and the errors of all landmarks are within 2 mm. We further tested the proposed model on 69 landmarks from the testing dataset that has a similar image pattern to the training pattern, resulting in a mean tracking error of 0.94+/-0.83 mm. Our experimental results have demonstrated the feasibility and accuracy of our proposed method in tracking liver anatomic landmarks using US images, providing a potential solution for real-time liver tracking for active motion management during radiation therapy.

Yuheng Li, Jacob Wynne, Jing Wang et al.

Biparametric magnetic resonance imaging (bpMRI) has demonstrated promising results in prostate cancer (PCa) detection using convolutional neural networks (CNNs). Recently, transformers have achieved competitive performance compared to CNNs in computer vision. Large scale transformers need abundant annotated data for training, which are difficult to obtain in medical imaging. Self-supervised learning (SSL) utilizes unlabeled data to generate meaningful semantic representations without the need for costly annotations, enhancing model performance on tasks with limited labeled data. We introduce a novel end-to-end Cross-Shaped windows (CSwin) transformer UNet model, CSwin UNet, to detect clinically significant prostate cancer (csPCa) in prostate bi-parametric MR imaging (bpMRI) and demonstrate the effectiveness of our proposed self-supervised pre-training framework. Using a large prostate bpMRI dataset with 1500 patients, we first pretrain CSwin transformer using multi-task self-supervised learning to improve data-efficiency and network generalizability. We then finetune using lesion annotations to perform csPCa detection. Five-fold cross validation shows that self-supervised CSwin UNet achieves 0.888 AUC and 0.545 Average Precision (AP), significantly outperforming four comparable models (Swin UNETR, DynUNet, Attention UNet, UNet). Using a separate bpMRI dataset with 158 patients, we evaluate our method robustness to external hold-out data. Self-supervised CSwin UNet achieves 0.79 AUC and 0.45 AP, still outperforming all other comparable methods and demonstrating good generalization to external data.

Shaoyan Pan, Shao-Yuan Lo, Min Huang et al.

In this work, we propose an adversarial attack-based data augmentation method to improve the deep-learning-based segmentation algorithm for the delineation of Organs-At-Risk (OAR) in abdominal Computed Tomography (CT) to facilitate radiation therapy. We introduce Adversarial Feature Attack for Medical Image (AFA-MI) augmentation, which forces the segmentation network to learn out-of-distribution statistics and improve generalization and robustness to noises. AFA-MI augmentation consists of three steps: 1) generate adversarial noises by Fast Gradient Sign Method (FGSM) on the intermediate features of the segmentation network's encoder; 2) inject the generated adversarial noises into the network, intentionally compromising performance; 3) optimize the network with both clean and adversarial features. Experiments are conducted segmenting the heart, left and right kidney, liver, left and right lung, spinal cord, and stomach. We first evaluate the AFA-MI augmentation using nnUnet and TT-Vnet on the test data from a public abdominal dataset and an institutional dataset. In addition, we validate how AFA-MI affects the networks' robustness to the noisy data by evaluating the networks with added Gaussian noises of varying magnitudes to the institutional dataset. Network performance is quantitatively evaluated using Dice Similarity Coefficient (DSC) for volume-based accuracy. Also, Hausdorff Distance (HD) is applied for surface-based accuracy. On the public dataset, nnUnet with AFA-MI achieves DSC = 0.85 and HD = 6.16 millimeters (mm); and TT-Vnet achieves DSC = 0.86 and HD = 5.62 mm. AFA-MI augmentation further improves all contour accuracies up to 0.217 DSC score when tested on images with Gaussian noises. AFA-MI augmentation is therefore demonstrated to improve segmentation performance and robustness in CT multi-organ segmentation.

Tian Liu, Anwesha Basu, James Caverlee et al.

Semi-supervised few-shot learning (SSFSL) formulates real-world applications like ''auto-annotation'', as it aims to learn a model over a few labeled and abundant unlabeled examples to annotate the unlabeled ones. Despite the availability of powerful open-source Vision-Language Models (VLMs) and their pretraining data, the SSFSL literature largely neglects these open-source resources. In contrast, the related area few-shot learning (FSL) has already exploited them to boost performance. Arguably, to achieve auto-annotation in the real world, SSFSL should leverage such open-source resources. To this end, we start by applying established SSL methods to finetune a VLM. Counterintuitively, they significantly underperform FSL baselines. Our in-depth analysis reveals the root cause: VLMs produce rather ''flat'' distributions of softmax probabilities. This results in zero utilization of unlabeled data and weak supervision signals. We address this issue with embarrassingly simple techniques: classifier initialization and temperature tuning. They jointly increase the confidence scores of pseudo-labels, improving the utilization rate of unlabeled data, and strengthening supervision signals. Building on this, we propose: Stage-Wise Finetuning with Temperature Tuning (SWIFT), which enables existing SSL methods to effectively finetune a VLM on limited labeled data, abundant unlabeled data, and task-relevant but noisy data retrieved from the VLM's pretraining set. Extensive experiments on five SSFSL benchmarks show that SWIFT outperforms recent FSL and SSL methods by $\sim$5 accuracy points. SWIFT even rivals supervised learning, which finetunes VLMs with the unlabeled data being labeled with ground truth!

Zibo Zhao, Zeqiang Lai, Qingxiang Lin et al.

We present Hunyuan3D 2.0, an advanced large-scale 3D synthesis system for generating high-resolution textured 3D assets. This system includes two foundation components: a large-scale shape generation model -- Hunyuan3D-DiT, and a large-scale texture synthesis model -- Hunyuan3D-Paint. The shape generative model, built on a scalable flow-based diffusion transformer, aims to create geometry that properly aligns with a given condition image, laying a solid foundation for downstream applications. The texture synthesis model, benefiting from strong geometric and diffusion priors, produces high-resolution and vibrant texture maps for either generated or hand-crafted meshes. Furthermore, we build Hunyuan3D-Studio -- a versatile, user-friendly production platform that simplifies the re-creation process of 3D assets. It allows both professional and amateur users to manipulate or even animate their meshes efficiently. We systematically evaluate our models, showing that Hunyuan3D 2.0 outperforms previous state-of-the-art models, including the open-source models and closed-source models in geometry details, condition alignment, texture quality, and etc. Hunyuan3D 2.0 is publicly released in order to fill the gaps in the open-source 3D community for large-scale foundation generative models. The code and pre-trained weights of our models are available at: https://github.com/Tencent/Hunyuan3D-2

Han Liu, Tian Liu, Kai Huang

As urban logistics demand continues to grow, UAV delivery has become a key solution to improve delivery efficiency, reduce traffic congestion, and lower logistics costs. However, to fully leverage the potential of UAV delivery networks, efficient swarm scheduling and management are crucial. In this paper, we propose a real-time scheduling and management system based on the ``Airport-Unloading Station" model, aiming to bridge the gap between high-level scheduling algorithms and low-level execution systems. This system, acting as middleware, accurately translates the requirements from the scheduling layer into specific execution instructions, ensuring that the scheduling algorithms perform effectively in real-world environments. Additionally, we implement three collaborative scheduling schemes involving autonomous ground vehicles (AGVs), unmanned aerial vehicles (UAVs), and ground staff to further optimize overall delivery efficiency. Through extensive experiments, this study demonstrates the rationality and feasibility of the proposed management system, providing practical solution for the commercial application of UAVs delivery in urban. Code: https://github.com/chengji253/UAVDeliverySystem

Tian Liu, Xueyang Hu, Tao Shu

Due to the distributed nature of Federated Learning (FL), researchers have uncovered that FL is vulnerable to backdoor attacks, which aim at injecting a sub-task into the FL without corrupting the performance of the main task. Single-shot backdoor attack achieves high accuracy on both the main task and backdoor sub-task when injected at the FL model convergence. However, the early-injected single-shot backdoor attack is ineffective because: (1) the maximum backdoor effectiveness is not reached at injection because of the dilution effect from normal local updates; (2) the backdoor effect decreases quickly as the backdoor will be overwritten by the newcoming normal local updates. In this paper, we strengthen the early-injected single-shot backdoor attack utilizing FL model information leakage. We show that the FL convergence can be expedited if the client trains on a dataset that mimics the distribution and gradients of the whole population. Based on this observation, we proposed a two-phase backdoor attack, which includes a preliminary phase for the subsequent backdoor attack. In the preliminary phase, the attacker-controlled client first launches a whole population distribution inference attack and then trains on a locally crafted dataset that is aligned with both the gradient and inferred distribution. Benefiting from the preliminary phase, the later injected backdoor achieves better effectiveness as the backdoor effect will be less likely to be diluted by the normal model updates. Extensive experiments are conducted on MNIST dataset under various data heterogeneity settings to evaluate the effectiveness of the proposed backdoor attack. Results show that the proposed backdoor outperforms existing backdoor attacks in both success rate and longevity, even when defense mechanisms are in place.

Pardis Taghavi, Tian Liu, Renjie Li et al.

Foundation models deliver strong perception but are often too computationally heavy to deploy, and adapting them typically requires costly annotations. We introduce a semi-supervised knowledge distillation (SSKD) framework that compresses pre-trained vision foundation models (VFMs) into compact experts using limited labeled and abundant unlabeled data, and instantiate it for instance segmentation where per-pixel labels are particularly expensive. The framework unfolds in three stages: (1) domain adaptation of the VFM(s) via self-training with contrastive calibration, (2) knowledge transfer through a unified multi-objective loss, and (3) student refinement to mitigate residual pseudo-label bias. Central to our approach is an instance-aware pixel-wise contrastive loss that fuses mask and class scores to extract informative negatives and enforce clear inter-instance margins. By maintaining this contrastive signal across both adaptation and distillation, we align teacher and student embeddings and more effectively leverage unlabeled images. On Cityscapes and ADE20K, our $\approx 11\times$ smaller student improves over its zero-shot VFM teacher(s) by +11.9 and +8.6 AP, surpasses adapted teacher(s) by +3.4 and +1.5 AP, and outperforms state-of-the-art SSKD methods on benchmarks.

Shubham Parashar, Zhiqiu Lin, Tian Liu et al.

Vision-language models (VLMs) excel in zero-shot recognition but their performance varies greatly across different visual concepts. For example, although CLIP achieves impressive accuracy on ImageNet (60-80%), its performance drops below 10% for more than ten concepts like night snake, presumably due to their limited presence in the pretraining data. However, measuring the frequency of concepts in VLMs' large-scale datasets is challenging. We address this by using large language models (LLMs) to count the number of pretraining texts that contain synonyms of these concepts. Our analysis confirms that popular datasets, such as LAION, exhibit a long-tailed concept distribution, yielding biased performance in VLMs. We also find that downstream applications of VLMs, including visual chatbots (e.g., GPT-4V) and text-to-image models (e.g., Stable Diffusion), often fail to recognize or generate images of rare concepts identified by our method. To mitigate the imbalanced performance of zero-shot VLMs, we propose REtrieval-Augmented Learning (REAL). First, instead of prompting VLMs using the original class names, REAL uses their most frequent synonyms found in pretraining texts. This simple change already outperforms costly human-engineered and LLM-enriched prompts over nine benchmark datasets. Second, REAL trains a linear classifier on a small yet balanced set of pretraining data retrieved using concept synonyms. REAL surpasses the previous zero-shot SOTA, using 400x less storage and 10,000x less training time!

Team Hunyuan3D, Shuhui Yang, Mingxin Yang et al.

3D AI-generated content (AIGC) is a passionate field that has significantly accelerated the creation of 3D models in gaming, film, and design. Despite the development of several groundbreaking models that have revolutionized 3D generation, the field remains largely accessible only to researchers, developers, and designers due to the complexities involved in collecting, processing, and training 3D models. To address these challenges, we introduce Hunyuan3D 2.1 as a case study in this tutorial. This tutorial offers a comprehensive, step-by-step guide on processing 3D data, training a 3D generative model, and evaluating its performance using Hunyuan3D 2.1, an advanced system for producing high-resolution, textured 3D assets. The system comprises two core components: the Hunyuan3D-DiT for shape generation and the Hunyuan3D-Paint for texture synthesis. We will explore the entire workflow, including data preparation, model architecture, training strategies, evaluation metrics, and deployment. By the conclusion of this tutorial, you will have the knowledge to finetune or develop a robust 3D generative model suitable for applications in gaming, virtual reality, and industrial design.

Zeqiang Lai, Yunfei Zhao, Haolin Liu et al.

In this report, we present Hunyuan3D 2.5, a robust suite of 3D diffusion models aimed at generating high-fidelity and detailed textured 3D assets. Hunyuan3D 2.5 follows two-stages pipeline of its previous version Hunyuan3D 2.0, while demonstrating substantial advancements in both shape and texture generation. In terms of shape generation, we introduce a new shape foundation model -- LATTICE, which is trained with scaled high-quality datasets, model-size, and compute. Our largest model reaches 10B parameters and generates sharp and detailed 3D shape with precise image-3D following while keeping mesh surface clean and smooth, significantly closing the gap between generated and handcrafted 3D shapes. In terms of texture generation, it is upgraded with phyiscal-based rendering (PBR) via a novel multi-view architecture extended from Hunyuan3D 2.0 Paint model. Our extensive evaluation shows that Hunyuan3D 2.5 significantly outperforms previous methods in both shape and end-to-end texture generation.

HunyuanWorld Team, Zhenwei Wang, Yuhao Liu et al.

Creating immersive and playable 3D worlds from texts or images remains a fundamental challenge in computer vision and graphics. Existing world generation approaches typically fall into two categories: video-based methods that offer rich diversity but lack 3D consistency and rendering efficiency, and 3D-based methods that provide geometric consistency but struggle with limited training data and memory-inefficient representations. To address these limitations, we present HunyuanWorld 1.0, a novel framework that combines the best of both worlds for generating immersive, explorable, and interactive 3D scenes from text and image conditions. Our approach features three key advantages: 1) 360° immersive experiences via panoramic world proxies; 2) mesh export capabilities for seamless compatibility with existing computer graphics pipelines; 3) disentangled object representations for augmented interactivity. The core of our framework is a semantically layered 3D mesh representation that leverages panoramic images as 360° world proxies for semantic-aware world decomposition and reconstruction, enabling the generation of diverse 3D worlds. Extensive experiments demonstrate that our method achieves state-of-the-art performance in generating coherent, explorable, and interactive 3D worlds while enabling versatile applications in virtual reality, physical simulation, game development, and interactive content creation.

Xiao Xiao, Yu Su, Sijing Zhang et al.

Large language models (LLMs) exhibit probabilistic output characteristics, yet conventional evaluation frameworks rely on deterministic scalar metrics. This study introduces a Bayesian approach for LLM capability assessment that integrates prior knowledge through probabilistic inference, addressing limitations under limited-sample regimes. By treating model capabilities as latent variables and leveraging a curated query set to induce discriminative responses, we formalize model ranking as a Bayesian hypothesis testing problem over mutually exclusive capability intervals. Experimental evaluations with GPT-series models demonstrate that the proposed method achieves superior discrimination compared to conventional evaluation methods. Results indicate that even with reduced sample sizes, the approach maintains statistical robustness while providing actionable insights, such as probabilistic statements about a model's likelihood of surpassing specific baselines. This work advances LLM evaluation methodologies by bridging Bayesian inference with practical constraints in real-world deployment scenarios.

Biwen Lei, Yang Li, Xinhai Liu et al.

The creation of high-quality 3D assets, a cornerstone of modern game development, has long been characterized by labor-intensive and specialized workflows. This paper presents Hunyuan3D Studio, an end-to-end AI-powered content creation platform designed to revolutionize the game production pipeline by automating and streamlining the generation of game-ready 3D assets. At its core, Hunyuan3D Studio integrates a suite of advanced neural modules (such as Part-level 3D Generation, Polygon Generation, Semantic UV, etc.) into a cohesive and user-friendly system. This unified framework allows for the rapid transformation of a single concept image or textual description into a fully-realized, production-quality 3D model complete with optimized geometry and high-fidelity PBR textures. We demonstrate that assets generated by Hunyuan3D Studio are not only visually compelling but also adhere to the stringent technical requirements of contemporary game engines, significantly reducing iteration time and lowering the barrier to entry for 3D content creation. By providing a seamless bridge from creative intent to technical asset, Hunyuan3D Studio represents a significant leap forward for AI-assisted workflows in game development and interactive media.

Raghavv Goel, Sudhanshu Agrawal, Mukul Gagrani et al.

In this paper, we introduce a simple training-free technique to improve the performance of drafter-based speculative decoding (SpD) methods that incorporates language modeling head (LM head) during drafting process. A drafter-based speculative decoding leverages one or more smaller language models, a.k.a. drafters or draft models, to sample a draft sequence or tree consisting of multiple tokens, followed by verification by a base LLM, a target model, accepting a subset as its valid generation. As it is usually considered that the speculative decoding requires one-to-one mapping between vocabularies of the target model and the draft model, it has been natural to share the vocabulary between them, or even share the LM head as in EAGLE or Medusa. We first identify that this draft token sampling scheme inherently contains an unnecessary inference overhead in drafting, especially for some target LLMs with very large vocabularies. Then, we propose a simple technique, VocabTrim, to mitigate the drafting overhead to improve the generation speed in memory-bound environment. VocabTrim reconstructs the drafter LM head to contain only a limited set of tokens, selected by the most frequently sampled from the vocabulary of the target model. While limiting the vocabulary in drafting slightly degrades the acceptance rate, it significantly reduces the drafting latency in memory-bound process which is often the case on edge devices, resulting in higher memory-bound speed up (MBSU). We show that our method can boost the memory-bound speed-up for Llama-3 models on Spec-Bench, specifically by 16% for Llama-3.2-3B-Instruct.

Ting-Ting Zhu, Yuan-Hai Shao, Chun-Na Li et al.

Learning using statistical invariants (LUSI) is a new learning paradigm, which adopts weak convergence mechanism, and can be applied to a wider range of classification problems. However, the computation cost of invariant matrices in LUSI is high for large-scale datasets during training. To settle this issue, this paper introduces a granularity statistical invariant for LUSI, and develops a new learning paradigm called learning using granularity statistical invariants (LUGSI). LUGSI employs both strong and weak convergence mechanisms, taking a perspective of minimizing expected risk. As far as we know, it is the first time to construct granularity statistical invariants. Compared to LUSI, the introduction of this new statistical invariant brings two advantages. Firstly, it enhances the structural information of the data. Secondly, LUGSI transforms a large invariant matrix into a smaller one by maximizing the distance between classes, achieving feasibility for large-scale datasets classification problems and significantly enhancing the training speed of model operations. Experimental results indicate that LUGSI not only exhibits improved generalization capabilities but also demonstrates faster training speed, particularly for large-scale datasets.

Tian Liu, Alex Cann, Ian Colbert et al.

While the rapid advancements in the reinforcement learning (RL) research community have been remarkable, the adoption in commercial video games remains slow. In this paper, we outline common challenges the Game AI community faces when using RL-driven NPCs in practice, and highlight the intersection of RL with traditional behavior trees (BTs) as a crucial juncture to be explored further. Although the BT+RL intersection has been suggested in several research papers, its adoption is rare. We demonstrate the viability of this approach using AMD Schola -- a plugin for training RL agents in Unreal Engine -- by creating multi-task NPCs in a complex 3D environment inspired by the commercial video game ``The Last of Us". We provide detailed methodologies for jointly training RL models with BTs while showcasing various skills.

Hanxin Wang, Tian Liu, Shu Kong

Pretrained VLMs achieve strong performance on downstream tasks when adapted with just a few labeled examples. As the adapted models inevitably encounter out-of-distribution (OOD) test data that deviates from the in-distribution (ID) task-specific training data, enhancing OOD generalization in few-shot adaptation is critically important. We study robust few-shot VLM adaptation, aiming to increase both ID and OOD accuracy. By comparing different adaptation methods (e.g., prompt tuning, linear probing, contrastive finetuning, and full finetuning), we uncover three key findings: (1) finetuning with proper hyperparameters significantly outperforms the popular VLM adaptation methods prompt tuning and linear probing; (2) visual encoder-only finetuning achieves better efficiency and accuracy than contrastively finetuning both visual and textual encoders; (3) finetuning the top layers of the visual encoder provides the best balance between ID and OOD accuracy. Building on these findings, we propose partial finetuning of the visual encoder empowered with two simple augmentation techniques: (1) retrieval augmentation which retrieves task-relevant data from the VLM's pretraining dataset to enhance adaptation, and (2) adversarial perturbation which promotes robustness during finetuning. Results show that the former/latter boosts OOD/ID accuracy while slightly sacrificing the ID/OOD accuracy. Yet, perhaps understandably, naively combining the two does not maintain their best OOD/ID accuracy. We address this dilemma with the developed SRAPF, Stage-wise Retrieval Augmentation-based Adversarial Partial Finetuning. SRAPF consists of two stages: (1) partial finetuning the visual encoder using both ID and retrieved data, and (2) adversarial partial finetuning with few-shot ID data. Extensive experiments demonstrate that SRAPF achieves the state-of-the-art ID and OOD accuracy on the ImageNet OOD benchmarks.

Pardis Taghavi, Tian Liu, Renjie Li et al.

Instance segmentation demands costly per-pixel annotations and computationally expensive models. We introduce CAST, a semi-supervised knowledge distillation (SSKD) framework that compresses pre-trained vision foundation models (VFM) into compact experts using limited labeled and abundant unlabeled data. CAST unfolds in three stages: (1) domain adaptation of the VFM(s) via self-training with contrastive calibration, (2) knowledge transfer through a unified multi-objective loss, and (3) student refinement to mitigate residual pseudo-label bias. Central to CAST is an \emph{instance-aware pixel-wise contrastive loss} that fuses mask and class scores to extract informative negatives and enforce clear inter-instance margins. By maintaining this contrastive signal across both adaptation and distillation, we align teacher and student embeddings and fully leverage unlabeled images. On Cityscapes and ADE20K, our ~11x smaller student improves over its zero-shot VFM teacher(s) by +8.5 and +7.1 AP, surpasses adapted teacher(s) by +3.4 and +1.5 AP, and further outperforms state-of-the-art SSKD methods on both benchmarks.

Jing Wang, Ruirui Liu, Yu Lei et al.

Purpose: Accurate 3D MRI-ultrasound (US) deformable registration is critical for real-time guidance in high-dose-rate (HDR) prostate brachytherapy. We present a weakly supervised spatial implicit neural representation (SINR) method to address modality differences and pelvic anatomy challenges. Methods: The framework uses sparse surface supervision from MRI/US segmentations instead of dense intensity matching. SINR models deformations as continuous spatial functions, with patient-specific surface priors guiding a stationary velocity field for biologically plausible deformations. Validation included 20 public Prostate-MRI-US-Biopsy cases and 10 institutional HDR cases, evaluated via Dice similarity coefficient (DSC), mean surface distance (MSD), and 95% Hausdorff distance (HD95). Results: The proposed method achieved robust registration. For the public dataset, prostate DSC was $0.93 \pm 0.05$, MSD $0.87 \pm 0.10$ mm, and HD95 $1.58 \pm 0.37$ mm. For the institutional dataset, prostate CTV achieved DSC $0.88 \pm 0.09$, MSD $1.21 \pm 0.38$ mm, and HD95 $2.09 \pm 1.48$ mm. Bladder and rectum performance was lower due to ultrasound's limited field of view. Visual assessments confirmed accurate alignment with minimal discrepancies. Conclusion: This study introduces a novel weakly supervised SINR-based approach for 3D MRI-US deformable registration. By leveraging sparse surface supervision and spatial priors, it achieves accurate, robust, and computationally efficient registration, enhancing real-time image guidance in HDR prostate brachytherapy and improving treatment precision.

Tian Liu, Huixin Zhang, Shubham Parashar et al.

Few-shot recognition (FSR) aims to train a classification model with only a few labeled examples of each concept concerned by a downstream task, where data annotation cost can be prohibitively high. We develop methods to solve FSR by leveraging a pretrained Vision-Language Model (VLM). We particularly explore retrieval-augmented learning (RAL), which retrieves open data, e.g., the VLM's pretraining dataset, to learn models for better serving downstream tasks. RAL has been studied in zero-shot recognition but remains under-explored in FSR. Although applying RAL to FSR may seem straightforward, we observe interesting and novel challenges and opportunities. First, somewhat surprisingly, finetuning a VLM on a large amount of retrieved data underperforms state-of-the-art zero-shot methods. This is due to the imbalanced distribution of retrieved data and its domain gaps with the few-shot examples in the downstream task. Second, more surprisingly, we find that simply finetuning a VLM solely on few-shot examples significantly outperforms previous FSR methods, and finetuning on the mix of retrieved and few-shot data yields even better results. Third, to mitigate the imbalanced distribution and domain gap issues, we propose Stage-Wise retrieval-Augmented fineTuning (SWAT), which involves end-to-end finetuning on mixed data in the first stage and retraining the classifier on the few-shot data in the second stage. Extensive experiments on nine popular benchmarks demonstrate that SWAT significantly outperforms previous methods by >6% accuracy.

Tian Liu, Yue Cui, Xueyang Hu et al.

Gossip learning (GL), as a decentralized alternative to federated learning (FL), is more suitable for resource-constrained wireless networks, such as Flying Ad-Hoc Networks (FANETs) that are formed by unmanned aerial vehicles (UAVs). GL can significantly enhance the efficiency and extend the battery life of UAV networks. Despite the advantages, the performance of GL is strongly affected by data distribution, communication speed, and network connectivity. However, how these factors influence the GL convergence is still unclear. Existing work studied the convergence of GL based on a virtual quantity for the sake of convenience, which failed to reflect the real state of the network when some nodes are inaccessible. In this paper, we formulate and investigate the impact of inaccessible nodes to GL under a dynamic network topology. We first decompose the weight divergence by whether the node is accessible or not. Then, we investigate the GL convergence under the dynamic of node accessibility and theoretically provide how the number of inaccessible nodes, data non-i.i.d.-ness, and duration of inaccessibility affect the convergence. Extensive experiments are carried out in practical settings to comprehensively verify the correctness of our theoretical findings.

Ming Hu, Tian Liu, Zhiwei Ling et al.

As a promising distributed machine learning paradigm, Federated Learning (FL) enables all the involved devices to train a global model collaboratively without exposing their local data privacy. However, for non-IID scenarios, the classification accuracy of FL models decreases drastically due to the weight divergence caused by data heterogeneity. Although various FL variants have been studied to improve model accuracy, most of them still suffer from the problem of non-negligible communication and computation overhead. In this paper, we introduce a novel FL approach named Fed-Cat that can achieve high model accuracy based on our proposed device selection strategy and device concatenation-based local training method. Unlike conventional FL methods that aggregate local models trained on individual devices, FedCat periodically aggregates local models after their traversals through a series of logically concatenated devices, which can effectively alleviate the model weight divergence problem. Comprehensive experimental results on four well-known benchmarks show that our approach can significantly improve the model accuracy of state-of-the-art FL methods without causing extra communication overhead.

Tian Liu, Jiahao Ding, Ting Wang et al.

As a promising method of central model training on decentralized device data while securing user privacy, Federated Learning (FL)is becoming popular in Internet of Things (IoT) design. However, when the data collected by IoT devices are highly skewed in a non-independent and identically distributed (non-IID) manner, the accuracy of vanilla FL method cannot be guaranteed. Although there exist various solutions that try to address the bottleneck of FL with non-IID data, most of them suffer from extra intolerable communication overhead and low model accuracy. To enable fast and accurate FL, this paper proposes a novel data-based device grouping approach that can effectively reduce the disadvantages of weight divergence during the training of non-IID data. However, since our grouping method is based on the similarity of extracted feature maps from IoT devices, it may incur additional risks of privacy exposure. To solve this problem, we propose an improved version by exploiting similarity information using the Locality-Sensitive Hashing (LSH) algorithm without exposing extracted feature maps. Comprehensive experimental results on well-known benchmarks show that our approach can not only accelerate the convergence rate, but also improve the prediction accuracy for FL with non-IID data.

Tian Liu, Zhiwei Ling, Jun Xia et al.

As a promising distributed machine learning paradigm, Federated Learning (FL) trains a central model with decentralized data without compromising user privacy, which has made it widely used by Artificial Intelligence Internet of Things (AIoT) applications. However, the traditional FL suffers from model inaccuracy since it trains local models using hard labels of data and ignores useful information of incorrect predictions with small probabilities. Although various solutions try to tackle the bottleneck of the traditional FL, most of them introduce significant communication and memory overhead, making the deployment of large-scale AIoT devices a great challenge. To address the above problem, this paper presents a novel Distillation-based Federated Learning (DFL) architecture that enables efficient and accurate FL for AIoT applications. Inspired by Knowledge Distillation (KD) that can increase the model accuracy, our approach adds the soft targets used by KD to the FL model training, which occupies negligible network resources. The soft targets are generated by local sample predictions of each AIoT device after each round of local training and used for the next round of model training. During the local training of DFL, both soft targets and hard labels are used as approximation objectives of model predictions to improve model accuracy by supplementing the knowledge of soft targets. To further improve the performance of our DFL model, we design a dynamic adjustment strategy for tuning the ratio of two loss functions used in KD, which can maximize the use of both soft targets and hard labels. Comprehensive experimental results on well-known benchmarks show that our approach can significantly improve the model accuracy of FL with both Independent and Identically Distributed (IID) and non-IID data.

Shunbo Zhang, Shun Zhang, Jianpeng Ma et al.

In a time-varying massive multiple-input multipleoutput (MIMO) system, the acquisition of the downlink channel state information at the base station (BS) is a very challenging task due to the prohibitively high overheads associated with downlink training and uplink feedback. In this paper, we consider the hybrid precoding structure at BS and examine the antennatime domain channel extrapolation. We design a latent ordinary differential equation (ODE)-based network under the variational auto-encoder (VAE) framework to learn the mapping function from the partial uplink channels to the full downlink ones at the BS side. Specifically, the gated recurrent unit is adopted for the encoder and the fully-connected neural network is used for the decoder. The end-to-end learning is utilized to optimize the network parameters. Simulation results show that the designed network can efficiently infer the full downlink channels from the partial uplink ones, which can significantly reduce the channel training overhead.

Qinbing Fu, Tian Liu, Xuelong Sun et al.

Visual dynamic complexity is a ubiquitous, hidden attribute of the visual world that every dynamic vision system is faced with. However, it is implicit and intractable which has never been quantitatively described due to the difficulty in defending temporal features correlated to spatial image complexity. To fill this vacancy, we propose a novel bio-robotic approach to profile visual dynamic complexity which can be used as a new metric. Here we apply a state-of-the-art brain-inspired motion detection neural network model to explicitly profile such complexity associated with spatial-temporal frequency (SF-TF) of visual scene. This model is for the first time implemented in an autonomous micro-mobile robot which navigates freely in an arena with visual walls displaying moving sine-wave grating or cluttered natural scene. The neural dynamic response can make reasonable prediction on surrounding complexity since it can be mapped monotonically to varying SF-TF of visual scene. The experiments show this approach is flexible to different visual scenes for profiling the dynamic complexity. We also use this metric as a predictor to investigate the boundary of another collision detection visual system performing in changing environment with increasing dynamic complexity. This research demonstrates a new paradigm of using biologically plausible visual processing scheme to estimate dynamic complexity of visual scene from both spatial and temporal perspectives, which could be beneficial to predicting input complexity when evaluating dynamic vision systems.

Mingquan Lin, Jacob Wynne, Yang Lei et al.

Medical imaging is widely used in cancer diagnosis and treatment, and artificial intelligence (AI) has achieved tremendous success in various tasks of medical image analysis. This paper reviews AI-based tumor subregion analysis in medical imaging. We summarize the latest AI-based methods for tumor subregion analysis and their applications. Specifically, we categorize the AI-based methods by training strategy: supervised and unsupervised. A detailed review of each category is presented, highlighting important contributions and achievements. Specific challenges and potential AI applications in tumor subregion analysis are discussed.

Yunfei Song, Tian Liu, Tongquan Wei et al.

Along with the proliferation of Artificial Intelligence (AI) and Internet of Things (IoT) techniques, various kinds of adversarial attacks are increasingly emerging to fool Deep Neural Networks (DNNs) used by Industrial IoT (IIoT) applications. Due to biased training data or vulnerable underlying models, imperceptible modifications on inputs made by adversarial attacks may result in devastating consequences. Although existing methods are promising in defending such malicious attacks, most of them can only deal with limited existing attack types, which makes the deployment of large-scale IIoT devices a great challenge. To address this problem, we present an effective federated defense approach named FDA3 that can aggregate defense knowledge against adversarial examples from different sources. Inspired by federated learning, our proposed cloud-based architecture enables the sharing of defense capabilities against different attacks among IIoT devices. Comprehensive experimental results show that the generated DNNs by our approach can not only resist more malicious attacks than existing attack-specific adversarial training methods, but also can prevent IIoT applications from new attacks.

Yang Lei, Yabo Fu, Tonghe Wang et al.

This paper presents a review of deep learning (DL) in multi-organ segmentation. We summarized the latest DL-based methods for medical image segmentation and applications. These methods were classified into six categories according to their network design. For each category, we listed the surveyed works, highlighted important contributions and identified specific challenges. Following the detailed review of each category, we briefly discussed its achievements, shortcomings and future potentials. We provided a comprehensive comparison among DL-based methods for thoracic and head & neck multiorgan segmentation using benchmark datasets, including the 2017 AAPM Thoracic Auto-segmentation Challenge datasets and 2015 MICCAI Head Neck Auto-Segmentation Challenge datasets.

Tonghe Wang, Yang Lei, Yabo Fu et al.

This paper reviewed the machine learning-based studies for quantitative positron emission tomography (PET). Specifically, we summarized the recent developments of machine learning-based methods in PET attenuation correction and low-count PET reconstruction by listing and comparing the proposed methods, study designs and reported performances of the current published studies with brief discussion on representative studies. The contributions and challenges among the reviewed studies were summarized and highlighted in the discussion part followed by.

Yabo Fu, Yang Lei, Tonghe Wang et al.

This paper presents a review of deep learning (DL) based medical image registration methods. We summarized the latest developments and applications of DL-based registration methods in the medical field. These methods were classified into seven categories according to their methods, functions and popularity. A detailed review of each category was presented, highlighting important contributions and identifying specific challenges. A short assessment was presented following the detailed review of each category to summarize its achievements and future potentials. We provided a comprehensive comparison among DL-based methods for lung and brain deformable registration using benchmark datasets. Lastly, we analyzed the statistics of all the cited works from various aspects, revealing the popularity and future trend of development in medical image registration using deep learning.

Tian Liu, Lichun Wang, Shaofan Wang

Fusing low level and high level features is a widely used strategy to provide details that might be missing during convolution and pooling. Different from previous works, we propose a new fusion mechanism called FillIn which takes advantage of prior knowledge described with superpixel segmentation. According to the prior knowledge, the FillIn chooses small region on low level feature map to fill into high level feature map. By using the proposed fusion mechanism, the low level features have equal channels for some tiny region as high level features, which makes the low level features have relatively independent power to decide final semantic label. We demonstrate the effectiveness of our model on PASCAL VOC 2012, it achieves competitive test result based on DeepLabv3+ backbone and visualizations of predictions prove our fusion can let small objects represent and low level features have potential for segmenting small objects.

Tian Liu, Tao Shu

Artificial neural network (ANN) provides superior accuracy for nonlinear alternating current (AC) state estimation (SE) in smart grid over traditional methods. However, research has discovered that ANN could be easily fooled by adversarial examples. In this paper, we initiate a new study of adversarial false data injection (FDI) attack against AC SE with ANN: by injecting a deliberate attack vector into measurements, the attacker can degrade the accuracy of ANN SE while remaining undetected. We propose a population-based algorithm and a gradient-based algorithm to generate attack vectors. The performance of these algorithms is evaluated through simulations on IEEE 9-bus, 14-bus and 30-bus systems under various attack scenarios. Simulation results show that DE is more effective than SLSQP on all simulation cases. The attack examples generated by DE algorithm successfully degrade the ANN SE accuracy with high probability.

Xuelong Sun, Tian Liu, Cheng Hu et al.

In the last few decades we have witnessed how the pheromone of social insect has become a rich inspiration source of swarm robotics. By utilising the virtual pheromone in physical swarm robot system to coordinate individuals and realise direct/indirect inter-robot communications like the social insect, stigmergic behaviour has emerged. However, many studies only take one single pheromone into account in solving swarm problems, which is not the case in real insects. In the real social insect world, diverse behaviours, complex collective performances and flexible transition from one state to another are guided by different kinds of pheromones and their interactions. Therefore, whether multiple pheromone based strategy can inspire swarm robotics research, and inversely how the performances of swarm robots controlled by multiple pheromones bring inspirations to explain the social insects' behaviours will become an interesting question. Thus, to provide a reliable system to undertake the multiple pheromone study, in this paper, we specifically proposed and realised a multiple pheromone communication system called ColCOS$Φ$. This system consists of a virtual pheromone sub-system wherein the multiple pheromone is represented by a colour image displayed on a screen, and the micro-robots platform designed for swarm robotics applications. Two case studies are undertaken to verify the effectiveness of this system: one is the multiple pheromone based on an ant's forage and another is the interactions of aggregation and alarm pheromones. The experimental results demonstrate the feasibility of ColCOS$Φ$ and its great potential in directing swarm robotics and social insects research.

Cunjing Ge, Feifei Ma, Tian Liu et al.

Constrained counting is important in domains ranging from artificial intelligence to software analysis. There are already a few approaches for counting models over various types of constraints. Recently, hashing-based approaches achieve both theoretical guarantees and scalability, but still rely on solution enumeration. In this paper, a new probabilistic polynomial time approximate model counter is proposed, which is also a hashing-based universal framework, but with only satisfiability queries. A variant with a dynamic stopping criterion is also presented. Empirical evaluation over benchmarks on propositional logic formulas and SMT(BV) formulas shows that the approach is promising.