Recently, Vision-Language-Action models (VLA) have advanced robot imitation learning, but high data collection costs and limited demonstrations hinder generalization and current imitation learning methods struggle in out-of-distribution scenarios, especially for long-horizon tasks. A key challenge is how to mitigate compounding errors in imitation learning, which lead to cascading failures over extended trajectories. To address these challenges, we propose the Diffusion Trajectory-guided Policy (DTP) framework, which generates 2D trajectories through a diffusion model to guide policy learning for long-horizon tasks. By leveraging task-relevant trajectories, DTP provides trajectory-level guidance to reduce error accumulation. Our two-stage approach first trains a generative vision-language model to create diffusion-based trajectories, then refines the imitation policy using them. Experiments on the CALVIN benchmark show that DTP outperforms state-of-the-art baselines by 25% in success rate, starting from scratch without external pretraining. Moreover, DTP significantly improves real-world robot performance.

We study the problem of contextual combinatorial semi-bandits, where input contexts are mapped into subsets of size $m$ of a collection of $K$ possible actions. In each round, the learner observes the realized reward of the predicted actions. Motivated by prototypical applications of contextual bandits, we focus on the $s$-sparse regime where we assume that the sum of rewards is bounded by some value $s\ll K$. For example, in recommendation systems the number of products purchased by any customer is significantly smaller than the total number of available products. Our main result is for the $(ε,δ)$-PAC variant of the problem for which we design an algorithm that returns an $ε$-optimal policy with high probability using a sample complexity of $\tilde{O}((poly(K/m)+sm/ε^2) \log(|Π|/δ))$ where $Π$ is the underlying (finite) class and $s$ is the sparsity parameter. This bound improves upon known bounds for combinatorial semi-bandits whenever $s\ll K$, and in the regime where $s=O(1)$, the leading term is independent of $K$. Our algorithm is also computationally efficient given access to an ERM oracle for $Π$. Our framework generalizes the list multiclass classification problem with bandit feedback, which can be seen as a special case with binary reward vectors. In the special case of single-label classification corresponding to $s=m=1$, we prove an $O((K^7+1/ε^2)\log(|H|/δ))$ sample complexity bound, which improves upon recent results in this scenario. Additionally, we consider the regret minimization setting where data can be generated adversarially, and establish a regret bound of $\tilde O(|Π|+\sqrt{smT\log |Π|})$, extending the result of Erez et al. (2024) who consider the simpler single label classification setting.

This paper analyzes multi-step temporal difference (TD)-learning algorithms within the ``deadly triad'' scenario, characterized by linear function approximation, off-policy learning, and bootstrapping. In particular, we prove that $n$-step TD-learning algorithms converge to a solution as the sampling horizon $n$ increases sufficiently. The paper is divided into two parts. In the first part, we comprehensively examine the fundamental properties of their model-based deterministic counterparts, including projected value iteration, gradient descent algorithms, which can be viewed as prototype deterministic algorithms whose analysis plays a pivotal role in understanding and developing their model-free reinforcement learning counterparts. In particular, we prove that these algorithms converge to meaningful solutions when $n$ is sufficiently large. Based on these findings, in the second part, two $n$-step TD-learning algorithms are proposed and analyzed, which can be seen as the model-free reinforcement learning counterparts of the model-based deterministic algorithms.

In this paper, we introduce the Curse of Depth, a concept that highlights, explains, and addresses the recent observation in modern Large Language Models (LLMs) where nearly half of the layers are less effective than expected. We first confirm the wide existence of this phenomenon across the most popular families of LLMs such as Llama, Mistral, DeepSeek, and Qwen. Our analysis, theoretically and empirically, identifies that the underlying reason for the ineffectiveness of deep layers in LLMs is the widespread usage of Pre-Layer Normalization (Pre-LN). While Pre-LN stabilizes the training of Transformer LLMs, its output variance exponentially grows with the model depth, which undesirably causes the derivative of the deep Transformer blocks to be an identity matrix, and therefore barely contributes to the training. To resolve this training pitfall, we propose LayerNorm Scaling (LNS), which scales the variance of output of the layer normalization inversely by the square root of its depth. This simple modification mitigates the output variance explosion of deeper Transformer layers, improving their contribution. Across a wide range of model sizes (130M to 7B), our experiments show that LNS consistently outperforms previous normalization and scaling techniques in enhancing LLM pre-training performance. Moreover, this improvement seamlessly carries over to supervised fine-tuning. All these gains can be attributed to the fact that LayerNorm Scaling enables deeper layers to contribute more effectively during training. Our code is available at \href{https://github.com/lmsdss/LayerNorm-Scaling}{LayerNorm-Scaling}.

Street view imagery is extensively utilized in representation learning for urban visual environments, supporting various sustainable development tasks such as environmental perception and socio-economic assessment. However, it is challenging for existing image representations to specifically encode the dynamic urban environment (such as pedestrians, vehicles, and vegetation), the built environment (including buildings, roads, and urban infrastructure), and the environmental ambiance (such as the cultural and socioeconomic atmosphere) depicted in street view imagery to address downstream tasks related to the city. In this work, we propose an innovative self-supervised learning framework that leverages temporal and spatial attributes of street view imagery to learn image representations of the dynamic urban environment for diverse downstream tasks. By employing street view images captured at the same location over time and spatially nearby views at the same time, we construct contrastive learning tasks designed to learn the temporal-invariant characteristics of the built environment and the spatial-invariant neighborhood ambiance. Our approach significantly outperforms traditional supervised and unsupervised methods in tasks such as visual place recognition, socioeconomic estimation, and human-environment perception. Moreover, we demonstrate the varying behaviors of image representations learned through different contrastive learning objectives across various downstream tasks. This study systematically discusses representation learning strategies for urban studies based on street view images, providing a benchmark that enhances the applicability of visual data in urban science. The code is available at https://github.com/yonglleee/UrbanSTCL.

Recent studies have demonstrated that many layers are functionally redundant in large language models (LLMs), enabling model compression by removing these layers to reduce inference cost. While such approaches can improve efficiency, indiscriminate layer pruning often results in significant performance degradation. In this paper, we propose GRASP (Gradient-based Retention of Adaptive Singular Parameters), a novel compression framework that mitigates this issue by preserving sensitivity-aware singular values. Unlike direct layer pruning, GRASP leverages gradient-based attribution on a small calibration dataset to adaptively identify and retain critical singular components. By replacing redundant layers with only a minimal set of parameters, GRASP achieves efficient compression while maintaining strong performance with minimal overhead. Experiments across multiple LLMs show that GRASP consistently outperforms existing compression methods, achieving 90% of the original model's performance under 20% compression ratio.

While Large Language Models (LLMs) demonstrate remarkable capabilities in scientific tasks such as literature analysis and experimental design (e.g., accurately extracting key findings from papers or generating coherent experimental procedures), existing evaluation benchmarks primarily assess performance using rich contextual inputs. We introduce LiveIdeaBench, a comprehensive benchmark evaluating LLMs' scientific idea generation by assessing divergent thinking capabilities using single-keyword prompts. Drawing from Guilford's creativity theory, our benchmark employs a dynamic panel of state-of-the-art LLMs to assess generated ideas across five key dimensions: originality, feasibility, fluency, flexibility, and clarity. Through extensive experimentation with over 40 leading models across 1,180 keywords spanning 22 scientific domains, we reveal that the scientific idea generation capabilities measured by our benchmark, are poorly predicted by standard metrics of general intelligence. Our results demonstrate that models like QwQ-32B-preview achieve creative performance comparable to top-tier models such as claude-3.7-sonnet:thinking, despite significant gaps in their general intelligence scores. These findings highlight the need for specialized evaluation benchmarks for scientific idea generation and suggest that enhancing these idea generation capabilities in LLMs may require different training strategies than those used for improving general problem-solving abilities, potentially enabling a wider range of AI tools tailored for different stages of the scientific process.

Detecting bias in multimodal news requires models that reason over text--image pairs, not just classify text. In response, we present ViLBias, a VQA-style benchmark and framework for detecting and reasoning about bias in multimodal news. The dataset comprises 40,945 text--image pairs from diverse outlets, each annotated with a bias label and concise rationale using a two-stage LLM-as-annotator pipeline with hierarchical majority voting and human-in-the-loop validation. We evaluate Small Language Models (SLMs), Large Language Models (LLMs), and Vision--Language Models (VLMs) across closed-ended classification and open-ended reasoning (oVQA), and compare parameter-efficient tuning strategies. Results show that incorporating images alongside text improves detection accuracy by 3--5\%, and that LLMs/VLMs better capture subtle framing and text--image inconsistencies than SLMs. Parameter-efficient methods (LoRA/QLoRA/Adapters) recover 97--99\% of full fine-tuning performance with $<5\%$ trainable parameters. For oVQA, reasoning accuracy spans 52--79\% and faithfulness 68--89\%, both improved by instruction tuning; closed accuracy correlates strongly with reasoning ($r = 0.91$). ViLBias offers a scalable benchmark and strong baselines for multimodal bias detection and rationale quality.

In this paper, we introduce RoboMIND (Multi-embodiment Intelligence Normative Data for Robot Manipulation), a dataset containing 107k demonstration trajectories across 479 diverse tasks involving 96 object classes. RoboMIND is collected through human teleoperation and encompasses comprehensive robotic-related information, including multi-view observations, proprioceptive robot state information, and linguistic task descriptions. To ensure data consistency and reliability for imitation learning, RoboMIND is built on a unified data collection platform and a standardized protocol, covering four distinct robotic embodiments: the Franka Emika Panda, the UR5e, the AgileX dual-arm robot, and a humanoid robot with dual dexterous hands. Our dataset also includes 5k real-world failure demonstrations, each accompanied by detailed causes, enabling failure reflection and correction during policy learning. Additionally, we created a digital twin environment in the Isaac Sim simulator, replicating the real-world tasks and assets, which facilitates the low-cost collection of additional training data and enables efficient evaluation. To demonstrate the quality and diversity of our dataset, we conducted extensive experiments using various imitation learning methods for single-task settings and state-of-the-art Vision-Language-Action (VLA) models for multi-task scenarios. By leveraging RoboMIND, the VLA models achieved high manipulation success rates and demonstrated strong generalization capabilities. To the best of our knowledge, RoboMIND is the largest multi-embodiment teleoperation dataset collected on a unified platform, providing large-scale and high-quality robotic training data. Our project is at https://x-humanoid-robomind.github.io/.

Combinatorics studies how discrete objects can be counted, arranged, and combined under specified rules. Motivated by uncertainty in real-world data and decisions, modern set-theoretic formalisms such as fuzzy sets, neutrosophic sets, rough sets, soft sets, and plithogenic sets have been developed. In particular, neutrosophic sets model uncertainty by assigning to each element degrees of truth, indeterminacy, and falsity. In parallel, these uncertainty frameworks are increasingly investigated in graphized and hyperized forms, where generalized graph models encompass classical graphs, hypergraphs, and higher-order "superhyper" structures; related hyper- and superhyper-concepts also arise beyond graph theory. This book (Edition 2.0) surveys and consolidates recent developments at the intersection of combinatorics, uncertain sets, uncertain graphs, and hyper/superhyper frameworks, while introducing several new graph and set concepts. As representative contributions, we extend graph-theoretic notions via Neutrosophic Oversets, Neutrosophic Undersets, Neutrosophic Offsets, and the Nonstandard Real Set. The second edition adds newly introduced concepts, corrects typographical issues, and re-examines mathematical consistency, aiming to serve as a compact reference and a source of inspiration for further research.

Rahimi and Recht (2007) introduced the idea of decomposing positive definite shift-invariant kernels by randomly sampling from their spectral distribution for machine learning applications. This famous technique, known as Random Fourier Features (RFF), is in principle applicable to any such kernel whose spectral distribution can be identified and simulated. In practice, however, it is usually applied to the Gaussian kernel because of its simplicity, since its spectral distribution is also Gaussian. Clearly, simple spectral sampling formulas would be desirable for broader classes of kernels. In this paper, we show that the spectral distribution of positive definite isotropic kernels in $\mathbb{R}^{d}$ for all $d\geq1$ can be decomposed as a scale mixture of $α$-stable random vectors, and we identify the mixing distribution as a function of the kernel. This constructive decomposition provides a simple and ready-to-use spectral sampling formula for many multivariate positive definite shift-invariant kernels, including exponential power kernels, and generalized Cauchy kernels, as well as newly introduced kernels such as the generalized Matérn, Tricomi, and Fox $H$ kernels. In particular, we retrieve the fact that the spectral distributions of these kernels, which can only be explicited in terms of the Fox $H$ special function, are scale mixtures of the multivariate Gaussian distribution, along with an explicit mixing distribution formula. This result has broad applications for support vector machines, kernel ridge regression, Gaussian processes, and other kernel-based machine learning techniques for which the random Fourier features technique is applicable.

Record matching models typically output a real-valued matching score that is later consumed through thresholding, ranking, or human review. While fairness in record matching has mostly been assessed using binary decisions at a fixed threshold, such evaluations can miss systematic disparities in the entire score distribution and can yield conclusions that change with the chosen threshold. We introduce a threshold-independent notion of score bias that extends standard group-fairness criteria-demographic parity (DP), equal opportunity (EO), and equalized odds (EOD)-from binary outputs to score functions by integrating group-wise metric gaps over all thresholds. Using this metric, we empirically show that several state-of-the-art deep matchers can exhibit substantial score bias even when appearing fair at commonly used thresholds. To mitigate these disparities without retraining the underlying matcher, we propose two model-agnostic post-processing methods that only require score evaluations on an (unlabeled) calibration set. Calib targets DP by aligning minority/majority score distributions to a common Wasserstein barycenter via a quantile-based optimal-transport map, with finite-sample guarantees on both residual DP bias and score distortion. C-Calib extends this idea to label-dependent notions (EO/EOD) by performing barycenter alignment conditionally on an estimated label, and we characterize how its guarantees depend on both sample size and label-estimation error. Experiments on standard record-matching benchmarks and multiple neural matchers confirm that Calib and C-Calib substantially reduce score bias with minimal loss in accuracy.

Ontology embeddings map classes, roles, and individuals in ontologies into $\mathbb{R}^n$, and within $\mathbb{R}^n$ similarity between entities can be computed or new axioms inferred. For ontologies in the Description Logic $\mathcal{EL}^{++}$, several optimization-based embedding methods have been developed that explicitly generate models of an ontology. However, these methods suffer from some limitations; they do not distinguish between statements that are unprovable and provably false, and therefore they may use entailed statements as negatives. Furthermore, they do not utilize the deductive closure of an ontology to identify statements that are inferred but not asserted. We evaluated a set of embedding methods for $\mathcal{EL}^{++}$ ontologies, incorporating several modifications that aim to make use of the ontology deductive closure. In particular, we designed novel negative losses that account both for the deductive closure and different types of negatives and formulated evaluation methods for knowledge base completion. We demonstrate that our embedding methods improve over the baseline ontology embedding in the task of knowledge base or ontology completion.

The Straight-Through Estimator (STE) is the dominant method for training neural networks with discrete variables, enabling gradient-based optimisation by routing gradients through a differentiable surrogate. However, existing STE variants conflate two fundamentally distinct concerns: forward-pass stochasticity, which controls exploration and latent space utilisation, and backward-pass gradient dispersion i.e how learning signals are distributed across categories. We show that these concerns are qualitatively different and that tying them to a single temperature parameter leaves significant performance gains untapped. We propose Decoupled Straight-Through (Decoupled ST), a minimal modification that introduces separate temperatures for the forward pass ($τ_f$) and the backward pass ($τ_b$). This simple change enables independent tuning of exploration and gradient dispersion. Across three diverse tasks (Stochastic Binary Networks, Categorical Autoencoders, and Differentiable Logic Gate Networks), Decoupled ST consistently outperforms Identity STE, Softmax STE, and Straight-Through Gumbel-Softmax. Crucially, optimal $(τ_f, τ_b)$ configurations lie far off the diagonal $τ_f = τ_b$, confirming that the two concerns do require different answers and that single-temperature methods are fundamentally constrained.

The current gold standard for the study of human movement is the marker-based motion capture system that offers high precision but constrained by costs and controlled environments. Markerless pose estimation systems emerge as ecological alternatives, allowing unobtrusive data acquisition in natural settings. This study compares the performance of two popular markerless systems, OpenPose (OP) and DeepLabCut (DLC), in assessing locomotion. Forty healthy subjects walked along a 5 meters walkway equipped with four force platforms and a camera. Gait parameters were obtained using OP BODY 25 Pre-Trained model (OPPT), DLC Model Zoo full human Pre-Trained model (DLCPT) and DLC Custom-Trained model (DLCCT), then compared with those acquired from the force platforms as reference system. Our results demonstrated that DLCCT outperformed DLCPT and OPPT, highlighting the importance of leveraging DeepLabCut transfer learning to enhance the pose estimation performance with a custom-trained neural networks. Moreover, DLCCT, with the implementation of the DLC refinement function, offers the most promising markerless pose estimation solution for evaluating locomotion. Therefore, our data provide insights into the DLC training and refinement processes required to achieve optimal performance. This study offers perspectives for clinicians and practitioners seeking accurate low-cost methods for movement assessment beyond laboratory settings.

Spatiotemporal prediction is important in solving natural problems and processing video frames, especially in weather forecasting and human action recognition. Recent advances attempt to incorporate prior physical knowledge into the deep learning framework to estimate the unknown governing partial differential equations (PDEs) in complex dynamics, which have shown promising results in spatiotemporal prediction tasks. However, previous approaches only restrict neural network architectures or loss functions to acquire physical or PDE features, which decreases the representative capacity of a neural network. Meanwhile, the updating process of the physical state cannot be effectively estimated. To solve the problems mentioned above, we introduce a physical-guided neural network, which utilizes an adaptive second-order Runge-Kutta method with physical constraints to model the physical states more precisely. Furthermore, we propose a frequency-enhanced Fourier module to strengthen the model's ability to estimate the spatiotemporal dynamics. We evaluate our model on both spatiotemporal and video prediction tasks. The experimental results show that our model outperforms several state-of-the-art methods and performs the best in several spatiotemporal scenarios with a much smaller parameter count.

Unstructured pruning is a popular compression method for efficiently reducing model parameters. However, while it effectively decreases the number of parameters, it is commonly believed that unstructured pruning cannot shorten the computational critical path, i.e., the maximum number of layers traversed during forward propagation. In this paper, we study when and how unstructured pruning can yield structural effects. For rectifier-activated networks, we introduce the notion of neuron entropy, which quantifies the degree of nonlinearity utilization. We show that magnitude-based pruning naturally lowers this entropy, sometimes down to zero-entropy layers that become linearizable and can thus be removed. Building on this insight, we propose a method that leverages "unstructured" pruning to favor sparsity in low-entropy layers, enabling their complete removal. We validate the phenomenon across CNNs, Vision Transformers, and NLP models: unstructured pruning can induce effective layer removal with little or no performance degradation in over-parameterized networks.

Visual Question Answering (VQA) is a challenging task that requires the joint understanding of natural language and visual content. While early research primarily focused on recognizing objects and scene context, it often overlooked scene text-an essential source of explicit semantic information. This paper introduces \textbf{ViTextVQA} (\textbf{Vi}etnamese \textbf{Text}-based \textbf{V}isual \textbf{Q}uestion \textbf{A}nswering), the first large-scale Vietnamese dataset specializing in text-based VQA. The dataset contains \textbf{over 16,000} images and \textbf{over 50,000} question-answer pairs. To tackle this task efficiently, \textbf{ViTextBLIP-2} (Vietnamese Text-based Bootstrapped Language-Image Model via Fine-tuning) is proposed, a novel multimodal feature fusion method designed to optimize Vietnamese text-based VQA. Experiments with state-of-the-art models highlight the importance of token ordering in OCR text for answer generation, leading to significant performance improvements. The ViTextVQA dataset is publicly available for research purposes.

Multi-agent reinforcement learning (MARL) for cyber-physical vehicle systems usually requires a significantly long training time due to their inherent complexity. Furthermore, deploying the trained policies in the real world demands a feature-rich environment along with multiple physical embodied agents, which may not be feasible due to monetary, physical, energy, or safety constraints. This work seeks to address these pain points by presenting a mixed-reality (MR) digital twin (DT) framework capable of: (i) boosting training speeds by selectively scaling parallelized simulation workloads on-demand, and (ii) immersing the MARL policies across hybrid simulation-to-reality (sim2real) experiments. The viability and performance of the proposed framework are highlighted through two representative use cases, which cover cooperative as well as competitive classes of MARL problems. We study the effect of: (i) agent and environment parallelization on training time, and (ii) systematic domain randomization on zero-shot sim2real transfer, across both case studies. Results indicate up to 76.3% reduction in training time with the proposed parallelization scheme and sim2real gap as low as 2.9% using the proposed deployment method.

The Potsdam Textbook Corpus (PoTeC) is a naturalistic eye-tracking-while-reading corpus containing data from 75 participants reading 12 scientific texts. PoTeC is the first naturalistic eye-tracking-while-reading corpus that contains eye-movements from domain-experts as well as novices in a within-participant manipulation: It is based on a 2x2x2 fully-crossed factorial design which includes the participants' level of study and the participants' discipline of study as between-subject factors and the text domain as a within-subject factor. The participants' reading comprehension was assessed by a series of text comprehension questions and their domain knowledge was tested by text-independent background questions for each of the texts. The materials are annotated for a variety of linguistic features at different levels. We envision PoTeC to be used for a wide range of studies including but not limited to analyses of expert and non-expert reading strategies. The corpus and all the accompanying data at all stages of the preprocessing pipeline and all code used to preprocess the data are made available via GitHub: https://github.com/DiLi-Lab/PoTeC.

Accurately gauging the confidence level of Large Language Models' (LLMs) predictions is pivotal for their reliable application. However, LLMs are often uncalibrated inherently and elude conventional calibration techniques due to their proprietary nature and massive scale. In this work, we explore the potential of deriving confidence from the distribution of multiple randomly sampled model generations, via three measures of consistency. We perform an extensive evaluation across various open and closed-source models on nine reasoning datasets. Results show that consistency-based calibration methods outperform existing post-hoc approaches. Meanwhile, we find that factors such as intermediate explanations, model scaling, and larger sample sizes enhance calibration, while instruction-tuning makes calibration more difficult. Moreover, confidence scores obtained from consistency have the potential to enhance model performance. Finally, we offer practical guidance on choosing suitable consistency metrics for calibration, tailored to the characteristics of various LMs.

Inspired by empirical work in neuroscience for Bayesian approaches to brain function, we give a unified probabilistic account of various types of symbolic reasoning from data. We characterise them in terms of formal logic using the classical consequence relation, an empirical consequence relation, maximal consistent sets, maximal possible sets and maximum likelihood estimation. The theory gives new insights into reasoning towards human-like machine intelligence.

Improving data utilization, especially for imperfect data from task failures, is crucial for robotic manipulation due to the challenging, time-consuming, and expensive data collection process in the real world. Current imitation learning (IL) typically discards imperfect data, focusing solely on successful expert data. While reinforcement learning (RL) can learn from explorations and failures, the sim2real gap and its reliance on dense reward and online exploration make it difficult to apply effectively in real-world scenarios. In this work, we aim to conquer the challenge of leveraging imperfect data without the need for reward information to improve the model performance for robotic manipulation in an offline manner. Specifically, we introduce a Self-Supervised Data Filtering framework (SSDF) that combines expert and imperfect data to compute quality scores for failed trajectory segments. High-quality segments from the failed data are used to expand the training dataset. Then, the enhanced dataset can be used with any downstream policy learning method for robotic manipulation tasks. Extensive experiments on the ManiSkill2 benchmark built on the high-fidelity Sapien simulator and real-world robotic manipulation tasks using the Franka robot arm demonstrated that the SSDF can accurately expand the training dataset with high-quality imperfect data and improve the success rates for all robotic manipulation tasks.

Despite extensive study, the significance of sharpness -- the trace of the loss Hessian at local minima -- remains unclear. We investigate an alternative perspective: how sharpness relates to the geometric structure of neural representations, specifically representation compression, defined as how strongly neural activations concentrate under local input perturbations. We introduce three measures -- Local Volumetric Ratio (LVR), Maximum Local Sensitivity (MLS), and Local Dimensionality -- and derive upper bounds showing these are mathematically constrained by sharpness: flatter minima necessarily limit compression. We extend these bounds to reparametrization-invariant sharpness and introduce network-wide variants (NMLS, NVR) that provide tighter, more stable bounds than prior single-layer analyses. Empirically, we validate consistent positive correlations across feedforward, convolutional, and transformer architectures. Our results suggest that sharpness fundamentally quantifies representation compression, offering a principled resolution to contradictory findings on the sharpness-generalization relationship.

Raw depth images captured in indoor scenarios frequently exhibit extensive missing values due to the inherent limitations of the sensors and environments. For example, transparent materials frequently elude detection by depth sensors; surfaces may introduce measurement inaccuracies due to their polished textures, extended distances, and oblique incidence angles from the sensor. The presence of incomplete depth maps imposes significant challenges for subsequent vision applications, prompting the development of numerous depth completion techniques to mitigate this problem. Numerous methods excel at reconstructing dense depth maps from sparse samples, but they often falter when faced with extensive contiguous regions of missing depth values, a prevalent and critical challenge in indoor environments. To overcome these challenges, we design a novel two-branch end-to-end fusion network named RDFC-GAN, which takes a pair of RGB and incomplete depth images as input to predict a dense and completed depth map. The first branch employs an encoder-decoder structure, by adhering to the Manhattan world assumption and utilizing normal maps from RGB-D information as guidance, to regress the local dense depth values from the raw depth map. The other branch applies an RGB-depth fusion CycleGAN, adept at translating RGB imagery into detailed, textured depth maps while ensuring high fidelity through cycle consistency. We fuse the two branches via adaptive fusion modules named W-AdaIN and train the model with the help of pseudo depth maps. Comprehensive evaluations on NYU-Depth V2 and SUN RGB-D datasets show that our method significantly enhances depth completion performance particularly in realistic indoor settings.

Statistical learning and logical reasoning are two major fields of AI expected to be unified for human-like machine intelligence. Most existing work considers how to combine existing logical and statistical systems. However, there is no theory of inference so far explaining how basic approaches to statistical learning and logical reasoning stem from a common principle. Inspired by the fact that much empirical work in neuroscience suggests Bayesian (or probabilistic generative) approaches to brain function including learning and reasoning, we here propose a simple Bayesian model of logical reasoning and statistical learning. The theory is statistically correct as it satisfies Kolmogorov's axioms, is consistent with both Fenstad's representation theorem and maximum likelihood estimation and performs exact Bayesian inference with a linear-time complexity. The theory is logically correct as it is a data-driven generalisation of uncertain reasoning from consistency, possibility, inconsistency and impossibility. The theory is correct in terms of machine learning as its solution to generation and prediction tasks on the MNIST dataset is not only empirically reasonable but also theoretically correct against the K nearest neighbour method. We simply model how data causes symbolic knowledge in terms of its satisfiability in formal logic. Symbolic reasoning emerges as a result of the process of going the causality forwards and backwards. The forward and backward processes correspond to an interpretation and inverse interpretation in formal logic, respectively. The inverse interpretation differentiates our work from the mainstream often referred to as inverse entailment, inverse deduction or inverse resolution. The perspective gives new insights into learning and reasoning towards human-like machine intelligence.

Humans rarely perceive objects in isolation but interpret scenes through relationships among co-occurring elements. How such contextual knowledge is acquired without explicit supervision remains unclear. Here we combine human psychophysics experiments with computational modelling to study the emergence of contextual reasoning. Participants were exposed to novel objects embedded in naturalistic scenes that followed predefined contextual rules capturing global context, local context and crowding. After viewing short training videos, participants completed a "lift-the-flap" task in which a hidden object had to be inferred from the surrounding context under variations in size, resolution and spatial arrangement. Humans rapidly learned these contextual associations without labels or feedback and generalised robustly across contextual changes. We then introduce SeCo (Self-supervised learning for Context Reasoning), a biologically inspired model that learns contextual relationships from complex scenes. SeCo encodes targets and context with separate vision encoders and stores latent contextual priors in a learnable external memory module. Given contextual cues, the model retrieves likely object representations to infer hidden targets. SeCo outperforms state-of-the-art self-supervised learning approaches and predicts object placements most consistent with human behaviour, highlighting the central role of contextual associations in scene understanding.

A novel Face Pyramid Vision Transformer (FPVT) is proposed to learn a discriminative multi-scale facial representations for face recognition and verification. In FPVT, Face Spatial Reduction Attention (FSRA) and Dimensionality Reduction (FDR) layers are employed to make the feature maps compact, thus reducing the computations. An Improved Patch Embedding (IPE) algorithm is proposed to exploit the benefits of CNNs in ViTs (e.g., shared weights, local context, and receptive fields) to model lower-level edges to higher-level semantic primitives. Within FPVT framework, a Convolutional Feed-Forward Network (CFFN) is proposed that extracts locality information to learn low level facial information. The proposed FPVT is evaluated on seven benchmark datasets and compared with ten existing state-of-the-art methods, including CNNs, pure ViTs, and Convolutional ViTs. Despite fewer parameters, FPVT has demonstrated excellent performance over the compared methods. Project page is available at https://khawar-islam.github.io/fpvt/

Recent studies have shown that Deep Leaning models are susceptible to adversarial examples, which are data, in general images, intentionally modified to fool a machine learning classifier. In this paper, we present a multi-objective nested evolutionary algorithm to generate universal unrestricted adversarial examples in a black-box scenario. The unrestricted attacks are performed through the application of well-known image filters that are available in several image processing libraries, modern cameras, and mobile applications. The multi-objective optimization takes into account not only the attack success rate but also the detection rate. Experimental results showed that this approach is able to create a sequence of filters capable of generating very effective and undetectable attacks.

Editing Large language models (LLMs) with real-world, unstructured knowledge is essential for correcting and updating their internal parametric knowledge. In this work, we revisit the fundamental next-token prediction (NTP) as a candidate paradigm for unstructured editing. We identify Context Reliance as a critical failure mode of NTP-based approaches, where knowledge acquired from edited text becomes highly dependent on its preceding context, leading to recall failures when that context is absent during inference. This hypothesis is supported by our empirical validation that prepending context during inference recovers knowledge recall. We further theoretically demonstrate that Context Reliance is an inherent consequence of gradient-based optimization, which tends to bind acquired knowledge to a specific aggregated contextual representation. To address this, we propose a simple yet effective COntext-INdependent editing framework (COIN), encouraging model to focus on knowledge within local scope rather than memorizing contextual patterns. Evaluations show that COIN reduces Context Reliance by 45.2% and outperforms strong baselines by 23.6% in editing success rate, highlighting the vital role of mitigating Context Reliance for robust editing.