Label noise is a common problem in real-world datasets, affecting both model training and validation. Clean data are essential for achieving strong performance and ensuring reliable evaluation. While various techniques have been proposed to detect noisy labels, there is no clear consensus on optimal approaches. We perform a comprehensive benchmark of detection methods by decomposing them into three fundamental components: label agreement function, aggregation method, and information gathering approach (in-sample vs out-of-sample). This decomposition can be applied to many existing detection methods, and enables systematic comparison across diverse approaches. To fairly compare methods, we propose a unified benchmark task, detecting a fraction of training samples equal to the dataset's noise rate. We also introduce a novel metric: the false negative rate at this fixed operating point. We identify that in-sample information gathering using average probability aggregation combined with the logit margin as the label agreement function achieves the best results across most scenarios. Our findings provide practical guidance for designing new detection methods and selecting techniques for specific applications. Keywords: Noisy label detection, Noisy labels, Dataset cleaning, Data quality, Benchmark, Neural networks 1. Introduction Most supervised learning methods assume a perfectly labeled dataset. However, training data often contain incorrectly labeled instances. Even large, standard benchmark datasets, such as CIFAR, ImageNet, and MS-COCO, are known to have noisy labels [1, 2].

Differentiating cardiogenic pulmonary oedema (CPE) from non-cardiogenic and structurally normal lungs in lung ultrasound (LUS) videos remains challenging due to the high visual variability of non-cardiogenic inflammatory patterns (NCIP/ARDS-like), interstitial lung disease, and healthy lungs. This heterogeneity complicates automated classification as overlapping B-lines and pleural artefacts are common. We introduce ZACH-ViT (Zero-token Adaptive Compact Hierarchical Vision Transformer), a 0.25 M-parameter Vision Transformer variant that removes both positional embeddings and the [CLS] token, making it fully permutation-invariant and suitable for unordered medical image data. To enhance generalization, we propose ShuffleStrides Data Augmentation (SSDA), which permutes probe-view sequences and frame orders while preserving anatomical validity. ZACH-ViT was evaluated on 380 LUS videos from 95 critically ill patients against nine state-of-the-art baselines. Despite the heterogeneity of the non-cardiogenic group, ZACH-ViT achieved the highest validation and test ROC-AUC (0.80 and 0.79) with balanced sensitivity (0.60) and specificity (0.91), while all competing models collapsed to trivial classification. It trains 1.35x faster than Minimal ViT (0.62M parameters) with 2.5x fewer parameters, supporting real-time clinical deployment. These results show that aligning architectural design with data structure can outperform scale in small-data medical imaging.

Misinformation spreads across web platforms through billions of daily multimodal posts that combine text and images, overwhelming manual fact-checking capacity. Supervised detection models require domain-specific training data and fail to generalize across diverse manipulation tactics. We present MIRAGE, an inference-time, model-pluggable agentic framework that decomposes multimodal verification into four sequential modules: visual veracity assessment detects AI-generated images, cross-modal consistency analysis identifies out-of-context repurposing, retrieval-augmented factual checking grounds claims in web evidence through iterative question generation, and a calibrated judgment module integrates all signals. MIRAGE orchestrates vision-language model reasoning with targeted web retrieval, outputs structured and citation-linked rationales. On MMFakeBench validation set (1,000 samples), MIRAGE with GPT-4o-mini achieves 81.65% F1 and 75.1% accuracy, outperforming the strongest zero-shot baseline (GPT-4V with MMD-Agent at 74.0% F1) by 7.65 points while maintaining 34.3% false positive rate versus 97.3% for a judge-only baseline. Test set results (5,000 samples) confirm generalization with 81.44% F1 and 75.08% accuracy. Ablation studies show visual verification contributes 5.18 F1 points and retrieval-augmented reasoning contributes 2.97 points. Our results demonstrate that decomposed agentic reasoning with web retrieval can match supervised detector performance without domain-specific training, enabling misinformation detection across modalities where labeled data remains scarce.

Abstract-- Advances in open-vocabulary semantic mapping and object navigation have enabled robots to perform an informed search of their environment for an arbitrary object. However, such zero-shot object navigation is typically designed for simple queries with an object name like "television" or "blue rug". Here, we consider more complex free-text queries with spatial relationships, such as "find the remote on the table" while still leveraging robustness of a semantic map. We present DIV-Nav, a real-time navigation system that efficiently addresses this problem through a series of relaxations: i) Decomposing natural language instructions with complex spatial constraints into simpler object-level queries on a semantic map, ii) computing the Intersection of individual semantic belief maps to identify regions where all objects co-exist, and iii) V alidating the discovered objects against the original, complex spatial constrains via a L VLM. We further investigate how to adapt the frontier exploration objectives of online semantic mapping to such spatial search queries to more effectively guide the search process. Robots operating in human environments must interpret natural language commands that go beyond simple object identification. While a command like "find a chair" requires handling simple object classes only, real-world search instructions often specify spatial relationships: "go to the chair next to the desk," "find the towel in the bathroom," or "get the book on the nightstand."

Advertisement (Ad) video violation detection is critical for ensuring platform compliance, but existing methods struggle with precise temporal grounding, noisy annotations, and limited generalization. We propose RAVEN, a novel framework that integrates curriculum reinforcement learning with multimodal large language models (MLLMs) to enhance reasoning and cognitive capabilities for violation detection. RAVEN employs a progressive training strategy, combining precisely and coarsely annotated data, and leverages Group Relative Policy Optimization (GRPO) to develop emergent reasoning abilities without explicit reasoning annotations. Multiple hierarchical sophisticated reward mechanism ensures precise temporal grounding and consistent category prediction. Experiments on industrial datasets and public benchmarks show that RAVEN achieves superior performances in violation category accuracy and temporal interval localization. We also design a pipeline to deploy the RAVEN on the online Ad services, and online A/B testing further validates its practical applicability, with significant improvements in precision and recall. RAVEN also demonstrates strong generalization, mitigating the catastrophic forgetting issue associated with supervised fine-tuning.

The rapid development of LLMs has raised concerns about their potential misuse, leading to various watermarking schemes that typically offer high detectability. However, existing watermarking techniques often face trade-off between watermark detectability and generated text quality. In this paper, we introduce Learning to Watermark (LTW), a novel selective watermarking framework that leverages multi-objective optimization to effectively balance these competing goals. LTW features a lightweight network that adaptively decides when to apply the watermark by analyzing sentence embeddings, token entropy, and current watermarking ratio. Training of the network involves two specifically constructed loss functions that guide the model toward Pareto-optimal solutions, thereby harmonizing watermark detectability and text quality. By integrating LTW with two baseline watermarking methods, our experimental evaluations demonstrate that LTW significantly enhances text quality without compromising detectability. Our selective watermarking approach offers a new perspective for designing watermarks for LLMs and a way to preserve high text quality for watermarks. The code is publicly available at: https://github.com/fattyray/learning-to-watermark

Large language models (LLMs) often generate reasoning traces that appear coherent but rest on unsupported assumptions, leading to hallucinated conclusions. Prior work mainly addresses factual hallucinations or relies on post-hoc verification, leaving reasoning-induced hallucinations largely unaddressed. We propose Audit-of-Understanding (AoU), a framework that constrains inference to validated premises through three phases: (1) decomposing a query into candidate assumptions, (2) auditing their support, and (3) conditioning inference only on the validated subset. Formally, AoU is \emph{posterior-constrained inference}, connecting to selective prediction and rejection learning. Our contributions are threefold: (i) theoretical guarantees under perfect validation, (ii) excess-risk bounds under imperfect audits, and (iii) tractability analysis. Empirically, AoU improves both accuracy and faithfulness on GSM8K, MultiArith, and SVAMP, achieving up to +30% gains on GSM8K, +45% on MultiArith, and consistent +20--28% improvements on SVAMP over Chain-of-Thought, Self-Consistency, and CoT-Decoding. Code is available at https://anonymous.4open.science/r/audit-of-understanding-E28B.

Wear and tear detection in fleet and shared vehicle systems is a critical challenge, particularly in rental and car-sharing services, where minor damage, such as dents, scratches, and underbody impacts, often goes unnoticed or is detected too late. Currently, manual inspection methods are the default approach, but are labour-intensive and prone to human error. In contrast, state-of-the-art image-based methods are less reliable when the vehicle is moving, and they cannot effectively capture underbody damage due to limited visual access and spatial coverage. This work introduces a novel multi-modal architecture based on anomaly detection to address these issues. Sensors such as Inertial Measurement Units (IMUs) and microphones are integrated into a compact device mounted on the vehicle's windshield. This approach supports real-time damage detection while avoiding the need for highly resource-intensive sensors. We developed multiple variants of multi-modal autoencoder-based architectures and evaluated them against unimodal and state-of-the-art methods. Our multi-modal ensemble model with pooling achieved the highest performance, with a Receiver Operating Characteristic-Area Under Curve (ROC-AUC) of 92%, demonstrating its effectiveness in real-world applications. This approach can also be extended to other applications, such as improving automotive safety. It can integrate with airbag systems for efficient deployment and help autonomous vehicles by complementing other sensors in collision detection.

Recent studies show near-perfect language-based predictions of depression scores (R2 = .70), but these "Mirror" models rely on language responses directly from depression assessments to predict depression assessment scores. These methods suffer from criterion contamination that inflate prediction estimates. We compare "Mirror" models to "Non-Mirror" models, which use other external language to predict depression scores. 110 participants completed both structured diagnostic (Mirror condition) and life history (Non-Mirror condition) interviews. LLMs were prompted to predict diagnostic depression scores. As expected, Mirror models were near-perfect. However, Non-Mirror models also displayed prediction sizes considered large in psychology. Further, both Mirror and Non-Mirror predictions correlated with other questionnaire-based depression symptoms at similar sizes, suggesting bias in Mirror models. Topic modeling revealed different theme structures across model types. As language models for depression continue to evolve, incorporating Non-Mirror approaches may support more valid and clinically useful language-based AI applications in psychological assessment.

Despite the rapid advancement of LLM-based agents, the reliable evaluation of their safety and security remains a significant challenge. Existing rule-based or LLM-based evaluators often miss dangers in agents' step-by-step actions, overlook subtle meanings, fail to see how small issues compound, and get confused by unclear safety or security rules. To overcome this evaluation crisis, we introduce AgentAuditor, a universal, training-free, memory-augmented reasoning framework that empowers LLM evaluators to emulate human expert evaluators. AgentAuditor constructs an experiential memory by having an LLM adaptively extract structured semantic features (e.g., scenario, risk, behavior) and generate associated chain-of-thought reasoning traces for past interactions. A multi-stage, context-aware retrieval-augmented generation process then dynamically retrieves the most relevant reasoning experiences to guide the LLM evaluator's assessment of new cases. Moreover, we developed ASSEBench, the first benchmark designed to check how well LLM-based evaluators can spot both safety risks and security threats. ASSEBench comprises 2293 meticulously annotated interaction records, covering 15 risk types across 29 application scenarios. A key feature of ASSEBench is its nuanced approach to ambiguous risk situations, employing "Strict" and "Lenient" judgment standards. Experiments demonstrate that AgentAuditor not only consistently improves the evaluation performance of LLMs across all benchmarks but also sets a new state-of-the-art in LLM-as-a-judge for agent safety and security, achieving human-level accuracy. Our work is openly accessible at https://github.com/Astarojth/AgentAuditor.