Object Hallucination in large vision-language models (LVLMs), where models generate non-factual content about input images, remains a critical barrier to their reliability in real-world applications. Existing mitigation strategies can be categorized into training-based and training-free methods. Training-based methods often achieve strong performance but are costly, requiring extensive computational resources, large-scale data, and time-consuming fine-tuning. Training-free approaches are particularly appealing due to their efficiency. However, existing training-free methods either require multiple decoding rounds, which adds computational overhead, or modify internal states in a model-specific way that risks degrading pretrained knowledge. We propose Test-Time Hallucination Mitigation (TTH) method, a novel training-free method that addresses both limitations. TTH introduces a token-validator module, implemented as a zero-shot Multi-Modal Classifier (MMC), to generate auxiliary logits grounded in the input image. These logits are fused with the original LVLM outputs at the token level for object tokens selected from a candidate pool. An entropy-based weighting scheme is then applied to enable robust and accurate predictions. Extensive experiments across multiple LVLM families and diverse benchmarks demonstrate that TTH consistently improves accuracy and robustness, underscoring its generalizability and practical effectiveness. Code is released at https://github.com/Mehran-TAM/TTH

While large vision–language models (LVLMs) achieve strong performance on multimodal tasks, they frequently generate hallucinations—unfaithful outputs misaligned with the visual input. To address this issue, we introduce CIPHER (Counterfactual Image Perturbations for Hallucination Extraction and Removal), a training-free method that suppresses vision-induced hallucinations via lightweight feature-level correction. Unlike prior training-free approaches that primarily focus on text-induced hallucinations, CIPHER explicitly targets hallucinations arising from the visual modality.CIPHER operates in two phases. In the offline phase, we construct OHC-25K (Object-Hallucinated Counterfactuals, 25,000 samples), a counterfactual dataset consisting of diffusion-edited images that intentionally contradict the original ground-truth captions. We pair these edited images with the unchanged ground-truth captions and process them through an LVLM to extract hallucination-related representations. Contrasting these representations with those from authentic (image, caption) pairs reveals structured, systematic shifts spanning a low-rank subspace characterizing vision-induced hallucination.In the inference phase, CIPHER suppresses hallucinations by projecting intermediate hidden states away from this subspace. Experiments across multiple benchmarks show that CIPHER significantly reduces hallucination rates while preserving task performance, demonstrating the effectiveness of counterfactual visual perturbations for improving LVLM faithfulness.

Modern GPU clusters are widely used to train distributed deep learning (DL) jobs, where scheduling decisions directly affect both system efficiency and user experience. In practice, DL jobs often come with deadlines, and failing to meet them can have serious consequences for users. Designing effective schedulers for such environments is challenging: scheduling decisions are tightly coupled, the problem is NP-hard, and most existing approaches either optimize job selection or GPU allocation in isolation, while largely overlooking deadline awareness and preemption. In this paper, we introduce PriorityDRL, a deadline-aware deep reinforcement learning (DRL) scheduler, with one agent for job selection and another for GPU allocation, jointly optimizing scheduling decisions. The job selection agent prioritizes jobs based on deadline urgency and supports preemption, allowing the system to respond dynamically to urgent jobs. The GPU allocation agent assigns jobs to physically proximal GPUs, reducing communication costs and mitigating resource fragmentation. By coordinating these complementary decisions, PriorityDRL balances deadline satisfaction with overall scheduling efficiency. Extensive experiments on real-world GPU cluster traces show that PriorityDRL consistently reduces missed deadlines and deadline lateness, improves GPU utilization, and achieves competitive job completion times compared to state-of-the-art heuristic and DRL-based schedulers. These results demonstrate that explicitly accounting for deadlines while separating job selection from GPU allocation is crucial for efficient and reliable operation in modern, shared GPU clusters.

3D Vision-Language Foundation Models (VLFMs) have demonstrated strong generalization and zero-shot recognition capabilities in open-world point cloud processing tasks. However, their performance often degrades in practical scenarios where data are noisy, incomplete, or drawn from distributions that differ from the training data. To address this challenge, we propose Uni-Adapter, a novel training-free online test-time adaptation (TTA) strategy for 3D VLFMs based on dynamic prototype learning. Uni-Adapter maintains a 3D cache that stores class-specific cluster centers as prototypes, which are continuously updated to capture intra-class variability under heterogeneous data distributions. These dynamic prototypes serve as anchors for cache-based logit computation through similarity scoring. In parallel, a graph-based label smoothing module models inter-prototype similarities to enforce label consistency among related prototypes. Finally, predictions from the original 3D VLFM and the refined 3D cache are unified through entropy-weighted aggregation to ensure reliable adaptation. Without retraining, Uni-Adapter effectively mitigates distribution shifts and achieves state-of-the-art performance across diverse 3D benchmarks and multiple 3D VLFMs, improving performance on ModelNet-40C by 10.55%, ScanObjectNN-C by 8.26%, and ShapeNet-C by 4.49% over the source 3D VLFMs.

Video reasoning segmentation requires localizing objects across video frames from natural language expressions, often involving spatial reasoning and implicit references. Recent approaches leverage frozen large vision-language models (LVLMs) by extracting attention maps and using them as spatial priors for segmentation, enabling training-free grounding. However, these attention maps are optimized for text generation rather than spatial localization, often resulting in diffuse and ambiguous grounding signals. In this work, we introduce SteerSeg, a lightweight framework that identifies attention misalignment as the key bottleneck in attention-based grounding and proposes to steer attention at its source through input-level conditioning. SteerSeg combines learnable soft prompts with reasoning-guided Chain-of-Thought (CoT) prompting. The soft prompts reshape the attention distribution to produce more spatially concentrated maps, while CoT-derived attributes resolve ambiguity among similar objects by guiding attention toward the correct instance. The resulting attention maps are converted into point prompts across keyframes to guide a segmentation model, while candidate tracklets are ranked and selected using correlation-based scoring. Our approach freezes the LVLM and segmentation model parameters and learns only a small set of soft prompts, preserving the model's pretrained reasoning capabilities while significantly improving grounding. Despite being trained only on Ref-YouTube-VOS, SteerSeg generalizes well across diverse benchmarks, significantly improving the spatial grounding capability of LVLMs. Project page: https://steerseg.github.io

Optimizing deep neural networks to generalize effectively across diverse visual domains remains a key challenge in computer vision, especially in domain-generalizable person re-identification (ReID). The goal of domain-generalizable ReID is to develop robust deep learning (DL) models that are effective across both known (source) and unseen (target) domains. However, many top-performing ReID methods overfit to the source domain, impairing their generalization ability. Previous approaches have employed Instance Normalization (IN) with learnable parameters to generalize domains and eliminate source domain styles. Recently, some DL frameworks have adopted normalization techniques without learnable parameters. We critically examine non-parametric normalization techniques for optimizing the deep ReID model, emphasizing the advantages of using non-parametric instance normalization as a gating mechanism to extract style-independent features at various abstraction levels within both convolutional neural networks (CNNs) and Vision Transformers (ViT). Our framework offers strategic guidance on the optimal placement of non-parametric IN within the network architecture to ensure effective information flow management in subsequent layers. Additionally, we employ one-dimensional Batch Normalization (BN) without learnable parameters at deeper network levels to remove content-related biases from the source domain. Our integrated approach, termed DualNormNP, systematically optimizes the model’s capacity to generalize across varied domains. Comprehensive evaluations on multiple benchmark ReID datasets demonstrate that our approach surpasses current state-of-the-art ReID methods in terms of generalization performance. Code is available on Github: https://github.com/mdamranhossenbhuiyan/DualNormNP

Humanitarian organizations are rapidly expanding their use of data in the pursuit of operational gains in effectiveness and efficiency. Ethical risks, particularly from artificial intelligence (AI) data processing, are increasingly recognized yet inadequately addressed by current humanitarian data protection guidelines. This study reports on a scoping review that maps the range of ethical issues that have been raised in the academic literature regarding data processing of people affected by humanitarian crises. We systematically searched databases to identify peer-reviewed studies published since 2010. Data and findings were standardized, grouping ethical issues into the value categories of autonomy, beneficence, non-maleficence, and justice. The study protocol followed Arksey and O’Malley’s approach and PRISMA reporting guidelines. We identified 16,200 unique records and retained 218 relevant studies. Nearly one in three (n = 66) discussed technologies related to AI. Seventeen studies included an author from a lower-middle income country while four included an author from a low-income country. We identified 22 ethical issues which were then grouped along the four ethical value categories of autonomy, beneficence, non-maleficence, and justice. Slightly over half of included studies (n = 113) identified ethical issues based on real-world examples. The most-cited ethical issue (n = 134) was a concern for privacy in cases where personal or sensitive data might be inadvertently shared with third parties. Aside from AI, the technologies most frequently discussed in these studies included social media, crowdsourcing, and mapping tools. Studies highlight significant concerns that data processing in humanitarian contexts can cause additional harm, may not provide direct benefits, may limit affected populations’ autonomy, and can lead to the unfair distribution of scarce resources. The increase in AI tool deployment for humanitarian assistance amplifies these concerns. Urgent development of specific, comprehensive guidelines, training, and auditing methods is required to address these ethical challenges. Moreover, empirical research from low and middle-income countries, disproportionally affected by humanitarian crises, is vital to ensure inclusive and diverse perspectives. This research should focus on the ethical implications of both emerging AI systems, as well as established humanitarian data management practices. Not applicable.

Large language models (LLMs) increasingly power car assistants, enabling natural language interaction for tasks such as maintenance, troubleshooting, and operational guidance. While retrieval-augmented generation (RAG) improves grounding using vehicle manuals, evaluating response quality remains a key challenge. Traditional metrics like BLEU and ROUGE fail to capture critical aspects such as factual accuracy and information coverage. We propose GEAR, a fully automated, reference-based evaluation framework for car assistant systems. GEAR uses LLMs as evaluators to compare assistant responses against ground-truth counterparts, assessing coverage, correctness, and other dimensions of answer quality. To enable fine-grained evaluation, both responses are decomposed into key facts and labeled as essential, optional, or safety-critical using LLMs. The evaluator then determines which of these facts are correct and covered. Experiments show that GEAR aligns closely with human annotations, offering a scalable and reliable solution for evaluating car assistants.

Public cloud GPU clusters are increasingly used for distributed deep learning, making job scheduling critical for minimizing waiting and completion times. However, this scheduling problem is NP-hard, and current approaches typically separate job scheduling from GPU allocation, yielding suboptimal performance. While deep reinforcement learning (DRL)-based methods offer flexibility, they often overlook two key challenges: (1) focusing solely on total completion time while ignoring fairness in user waiting times, and (2) neglecting GPU communication costs that significantly impact distributed training speed. We introduce AttentiveSched, a DRL-based framework that simultaneously optimizes job selection and GPU assignment. AttentiveSched leverages cluster topology for informed scheduling decisions, employing attention-based agents to capture global relationships in input sequences. Through reward functions addressing fairness, completion time, and communication costs, AttentiveSched out-performs multiple heuristics-based, meta-heuristics-based, and DRL-based schedulers on real-world datasets.

Efficient data migration is essential for organizations facing growing data volumes, evolving infrastructure, and network cost constraints. We focus on two common migration scenarios: high-speed cloud onboarding under tight downtime limits, and bandwidth-constrained synchronization over cost-sensitive links. While recent frameworks like DAMOCRO improve throughput and reduce cost using offline-trained classifiers to group similar data for compression, they suffer from the workload imbalance issue and the use of a fixed compression level. In this paper, we propose two complementary improvements to DAMOCRO: (1) a balanced clustering-based classifier that ensures even workload distribution across compression workers, which helps reduce bottlenecks and improve scalability; and (2) an adaptive compression-level predictor that adaptively selects compression settings based on data characteristics and network conditions to optimize end-to-end throughput. Experiments on five datasets demonstrate up to 43.6 % throughput improvement and up to 23.7 % cost reduction, confirming the effectiveness of the proposed enhancements in diverse industrial data migration settings.

Answering “Where is the X button?” with “It’s next to the Y button” is unhelpful if the user knows neither location. Useful answers require obvious landmarks as a reference point. We address this by generating from a vehicle dashboard diagram a spatial knowledge graph (SKG) that shows the spatial relationship between a dashboard component and its nearby landmarks and using the SKG to help answer questions. We evaluate three distinct generation pipelines (Per-Attribute, Per-Component, and a Single-Shot baseline) to create the SKG using Large Vision-Language Models (LVLMs). On a new 65-vehicle dataset, we demonstrate that a decomposed Per-Component pipeline is the most effective strategy for generating a high-quality SKG; the graph produced by this method, when evaluated with a novel Significance score, identifies landmarks achieving 71.3% agreement with human annotators. This work enables downstream QA systems to provide more intuitive, landmark-based answers.

Pretrained vision-language models (VLMs) like CLIP show strong zero-shot performance but struggle with generalization under distribution shifts. Test-Time Adaptation (TTA) addresses this by adapting VLMs to unlabeled test data in new domains. While some TTA methods rely on prompt-tuning, training-free cache-based approaches are preferred for efficiency. However, current cache-based TTA models store only a limited set of high-confidence samples, restricting the decision boundary to these samples and ignoring the influence of other incoming test data. To address this, we propose Efficient Test-Time Adaptation (ETTA), introducing a Recursive Updating module that integrates all incoming test samples, progressively refining the decision boundary. This strategy mimics an unbounded cache, dynamically updating contextual embeddings for improved accuracy with minimal memory and computational overhead. ETTA also includes an Adaptive Ensemble module to reduce prompt dependency in image-to-text scores by dynamically selecting optimal prompts for each class. Furthermore, ETTA adaptively combines scores from both modules based on confidence levels, leveraging their complementary strengths. Extensive experiments on two benchmarks confirm that ETTA surpasses the state-of-the-art TTA models in computational complexity and accuracy, setting a new standard for effective, efficient test-time adaptation. The code has been released at https://github.com/hamidreza-dastmalchi/ETTA.

Person re-identification (ReID) has evolved from handcrafted feature-based methods to deep learning approaches and, more recently, to models incorporating large language models (LLMs). Early methods struggled with variations in lighting, pose, and viewpoint, but deep learning addressed these issues by learning robust visual features. Building on this, LLMs now enable ReID systems to integrate semantic and contextual information through natural language. This survey traces that full evolution and offers one of the first comprehensive reviews of ReID approaches that leverage LLMs, where textual descriptions are used as privileged information to improve visual matching. A key contribution is the use of dynamic, identity-specific prompts generated by GPT-4o, which enhance the alignment between images and text in vision-language ReID systems. Experimental results show that these descriptions improve accuracy, especially in complex or ambiguous cases. To support further research, we release a large set of GPT-4o-generated descriptions for standard ReID datasets. By bridging computer vision and natural language processing, this survey offers a unified perspective on the field's development and outlines key future directions such as better prompt design, cross-modal transfer learning, and real-world adaptability.

Test-time adaptation (TTA) of 3D point clouds is crucial for mitigating discrepancies between training and testing samples in real-world scenarios, particularly when handling corrupted point clouds. LiDAR data, for instance, can be affected by sensor failures or environmental factors, causing domain gaps. Adapting models to these distribution shifts online is crucial, as training for every possible variation is impractical. Existing methods often focus on fine-tuning pre-trained models based on self-supervised learning or pseudo-labeling, which can lead to forgetting valuable source domain knowledge over time and reduce generalization on future tests. In this paper, we introduce a novel 3D test-time adaptation method, termed 3DD-TTA, which stands for 3D Denoising Diffusion Test-Time Adaptation. This method uses a diffusion strategy that adapts input point cloud samples to the source domain while keeping the source model parameters intact. The approach uses a Variational Autoencoder (VAE) to encode the corrupted point cloud into a shape latent and latent points. These latent points are corrupted with Gaussian noise and subjected to a denoising diffusion process. During this process, both the shape latent and latent points are updated to preserve fidelity, guiding the denoising toward generating consistent samples that align more closely with the source domain. We conduct extensive experiments on the ShapeNet dataset and investigate its generalizability on ModelNet40 and ScanObjectNN, achieving state-of-the-art results. The code has been released at https://github.com/hamidreza-dastmalchi/3DD-Tta.

Infectious diseases occur when pathogens from other individuals or animals infect a person, causing harm to both individuals and society. Outbreaks of such diseases can pose a significant threat to human health. However, early detection and tracking of these outbreaks have the potential to reduce mortality rates. To address these threats, public health authorities have endeavored to establish comprehensive mechanisms for collecting disease data. Many countries have implemented infectious disease surveillance systems, with epidemic detection as a primary objective. The clinical healthcare system, local/state health agencies, federal agencies, academic/professional groups, and collaborating governmental entities all play pivotal roles within this system. Moreover, search engines and social media platforms can serve as valuable tools for monitoring disease trends. The Internet and social media have become significant platforms where users share information about their preferences and relationships. This real-time information can be harnessed to gauge the influence of ideas and societal opinions, proving highly useful across various domains and research areas, such as marketing campaigns, financial predictions, and public health. This article provides a review of the existing standard methods developed by researchers for detecting outbreaks using time series data. These methods leverage various data sources, including conventional data sources and social media data or Internet data sources. The review particularly concentrates on works published within the timeframe of 2015 to 2022.

Graphs representation learning has been a very active research area in recent years. The goal of graph representation learning is to generate graph representation vectors that capture the structure and features of large graphs accurately. This is especially important because the quality of the graph representation vectors will affect the performance of these vectors in downstream tasks such as node classification, link prediction and anomaly detection. Many techniques are proposed for generating effective graph representation vectors. Two of the most prevalent categories of graph representation learning are graph embedding methods without using graph neural nets (GNN), which we denote as non-GNN based graph embedding methods, and graph neural nets (GNN) based methods. Non-GNN graph embedding methods are based on techniques such as random walks, temporal point processes and neural network learning methods. GNN-based methods, on the other hand, are the application of deep learning on graph data. In this survey, we provide an overview of these two categories and cover the current state-of-the-art methods for both static and dynamic graphs. Finally, we explore some open and ongoing research directions for future work.

To enhance a question-answering system for automotive drivers, we tackle the problem of automatic generation of icon image descriptions. The descriptions can match the driver’s query about the icon appearing on the dashboard and tell the driver what is happening so that they may take an appropriate action. We use three state-of-the-art large vision-language models to generate both visual and functional descriptions based on the icon image and its context information in the car manual. Both zero-shot and few-shot prompts are used. We create a dataset containing over 400 icons with their ground-truth descriptions and use it to evaluate model-generated descriptions across several performance metrics. Our evaluation shows that two of these models (GPT-4o and Claude 3.5) performed well on this task, while the third model (LLaVA-NEXT) performs poorly.

Affective tasks, including sentiment analysis, emotion classification, and sarcasm detection have drawn a lot of attention in recent years due to a broad range of useful applications in various domains.The main goal of affect detection tasks is to recognize states such as mood, sentiment, and emotions from textual data (e.g., news articles or product reviews).Despite the importance of utilizing preprocessing steps in different stages (i.e., word representation learning and building a classification model) of affect detection tasks, this topic has not been studied well.To that end, we explore whether applying various preprocessing methods (stemming, lemmatization, stopword removal, punctuation removal and so on) and their combinations in different stages of the affect detection pipeline can improve the model performance.The are many preprocessing approaches that can be utilized in affect detection tasks.However, their influence on the final performance depends on the type of preprocessing and the stages that they are applied.Moreover, the preprocessing impacts vary across different affective tasks.Our analysis provides thorough insights into how preprocessing steps can be applied in building an effect detection pipeline and their respective influence on performance.