id
stringlengths
9
16
title
stringlengths
4
278
categories
stringlengths
5
104
abstract
stringlengths
6
4.09k
2502.10012
Dream to Drive: Model-Based Vehicle Control Using Analytic World Models
cs.AI cs.RO
Differentiable simulators have recently shown great promise for training autonomous vehicle controllers. Being able to backpropagate through them, they can be placed into an end-to-end training loop where their known dynamics turn into useful priors for the policy to learn, removing the typical black box assumption of the environment. So far, these systems have only been used to train policies. However, this is not the end of the story in terms of what they can offer. Here, for the first time, we use them to train world models. Specifically, we present three new task setups that allow us to learn next state predictors, optimal planners, and optimal inverse states. Unlike analytic policy gradients (APG), which requires the gradient of the next simulator state with respect to the current actions, our proposed setups rely on the gradient of the next state with respect to the current state. We call this approach Analytic World Models (AWMs) and showcase its applications, including how to use it for planning in the Waymax simulator. Apart from pushing the limits of what is possible with such simulators, we offer an improved training recipe that increases performance on the large-scale Waymo Open Motion dataset by up to 12% compared to baselines at essentially no additional cost.
2502.10013
Probabilistic Lexical Manifold Construction in Large Language Models via Hierarchical Vector Field Interpolation
cs.CL
Hierarchical vector field interpolation introduces a structured probabilistic framework for lexical representation, ensuring that word embeddings transition smoothly across a continuous manifold rather than being constrained to discrete token mappings. The proposed methodology constructs a probabilistic function space where word representations adhere to topological consistency, mitigating representational discontinuities commonly observed in transformer-based embeddings. Empirical evaluations reveal that probabilistic constraints enhance lexical coherence by refining contextual relationships, leading to improvements in semantic stability across multiple linguistic distributions. The application of divergence minimization techniques ensures that interpolated embeddings maintain probabilistic consistency while preserving computational feasibility for large-scale implementations. Experimental findings demonstrate that interpolated lexical manifolds improve representation density alignment, reducing anisotropic distortions in contextual embedding distributions. Comparative analyses with standard transformer-based models highlight that structured interpolation yields more stable representations, particularly in tasks requiring fine-grained semantic differentiation. The statistical evaluation of embedding divergence confirms that probabilistic lexical manifolds reduce representational inconsistencies while maintaining coherence across varying scales of contextual abstraction. An assessment of computational efficiency reveals that while interpolation introduces minor processing overhead, the structured representation learning approach remains scalable for practical deployment.
2502.10014
Recovering nonlinear dynamics from non-uniform observations: A physics-based identification approach with practical case studies
eess.SY cs.SY
Uniform and smooth data collection is often infeasible in real-world scenarios. In this paper, we propose an identification framework to effectively handle the so-called non-uniform observations, i.e., data scenarios that include missing measurements, multiple runs, or aggregated observations. The goal is to provide a general approach for accurately recovering the overall dynamics of possibly nonlinear systems, allowing the capture of the system behavior over time from non-uniform observations. The proposed approach exploits prior knowledge by integrating domain-specific, interpretable, physical principles with black-box approximators, proving significant flexibility and adaptability in handling different types of non-uniform measurements, and addressing the limitations of traditional linear and black-box methods. The description of this novel framework is supported by a theoretical study on the effect of non-uniform observations on the accuracy of parameter estimation. Specifically, we demonstrate the existence of upper bounds on the parametric error resulting from missing measurements and aggregated observations. Then, the effectiveness of the approach is demonstrated through two case studies. These include a practical application with missing samples, i.e., the identification of a continuous stirred-tank reactor using real data, and a simulated Lotka-Volterra system under aggregated observations. The results highlight the ability of the framework to robustly estimate the system parameters and to accurately reconstruct the model dynamics despite the availability of non-uniform measurements.
2502.10019
A Differential Equation Approach to the Most-Informative Boolean Function Conjecture
cs.IT math.IT
We study the most-informative Boolean function conjecture using a differential equation approach. This leads to a formulation of a functional inequality on finite-dimensional random variables. We also develop a similar inequality in the case of the Hellinger conjecture. Finally, we conjecture a specific finite dimensional inequality that, if proved, will lead to a proof of the Boolean function conjecture in the balanced case. We further show that the above inequality holds modulo four explicit inequalities (all of which seems to hold via numerical simulation) with the first three containing just two variables and a final one involving four variables.
2502.10020
Improved Online Confidence Bounds for Multinomial Logistic Bandits
stat.ML cs.LG
In this paper, we propose an improved online confidence bound for multinomial logistic (MNL) models and apply this result to MNL bandits, achieving variance-dependent optimal regret. Recently, Lee & Oh (2024) established an online confidence bound for MNL models and achieved nearly minimax-optimal regret in MNL bandits. However, their results still depend on the norm-boundedness of the unknown parameter $B$ and the maximum size of possible outcomes $K$. To address this, we first derive an online confidence bound of $O\left(\sqrt{d \log t} + B \right)$, which is a significant improvement over the previous bound of $O (B \sqrt{d} \log t \log K )$ (Lee & Oh, 2024). This is mainly achieved by establishing tighter self-concordant properties of the MNL loss and introducing a novel intermediary term to bound the estimation error. Using this new online confidence bound, we propose a constant-time algorithm, OFU-MNL++, which achieves a variance-dependent regret bound of $O \Big( d \log T \sqrt{ \smash[b]{\sum_{t=1}^T} \sigma_t^2 } \Big) $ for sufficiently large $T$, where $\sigma_t^2$ denotes the variance of the rewards at round $t$, $d$ is the dimension of the contexts, and $T$ is the total number of rounds. Furthermore, we introduce a Maximum Likelihood Estimation (MLE)-based algorithm, OFU-MN$^2$L, which achieves an anytime poly(B)-free regret of $O \Big( d \log (BT) \sqrt{ \smash[b]{\sum_{t=1}^T} \sigma_t^2 } \Big) $.
2502.10027
Heterogeneous Resource Allocation with Multi-task Learning for Wireless Networks
cs.LG
The optimal solution to an optimization problem depends on the problem's objective function, constraints, and size. While deep neural networks (DNNs) have proven effective in solving optimization problems, changes in the problem's size, objectives, or constraints often require adjustments to the DNN architecture to maintain effectiveness, or even retraining a new DNN from scratch. Given the dynamic nature of wireless networks, which involve multiple and diverse objectives that can have conflicting requirements and constraints, we propose a multi-task learning (MTL) framework to enable a single DNN to jointly solve a range of diverse optimization problems. In this framework, optimization problems with varying dimensionality values, objectives, and constraints are treated as distinct tasks. To jointly address these tasks, we propose a conditional computation-based MTL approach with routing. The multi-task DNN consists of two components, the base DNN (bDNN), which is the single DNN used to extract the solutions for all considered optimization problems, and the routing DNN (rDNN), which manages which nodes and layers of the bDNN to be used during the forward propagation of each task. The output of the rDNN is a binary vector which is multiplied with all bDNN's weights during the forward propagation, creating a unique computational path through the bDNN for each task. This setup allows the tasks to either share parameters or use independent ones, with the decision controlled by the rDNN. The proposed framework supports both supervised and unsupervised learning scenarios. Numerical results demonstrate the efficiency of the proposed MTL approach in solving diverse optimization problems. In contrast, benchmark DNNs lacking the rDNN mechanism were unable to achieve similar levels of performance, highlighting the effectiveness of the proposed architecture.
2502.10028
ManiTrend: Bridging Future Generation and Action Prediction with 3D Flow for Robotic Manipulation
cs.CV cs.RO
Language-conditioned manipulation is a vital but challenging robotic task due to the high-level abstraction of language. To address this, researchers have sought improved goal representations derived from natural language. In this paper, we highlight 3D flow - representing the motion trend of 3D particles within a scene - as an effective bridge between language-based future image generation and fine-grained action prediction. To this end, we develop ManiTrend, a unified framework that models the dynamics of 3D particles, vision observations and manipulation actions with a causal transformer. Within this framework, features for 3D flow prediction serve as additional conditions for future image generation and action prediction, alleviating the complexity of pixel-wise spatiotemporal modeling and providing seamless action guidance. Furthermore, 3D flow can substitute missing or heterogeneous action labels during large-scale pretraining on cross-embodiment demonstrations. Experiments on two comprehensive benchmarks demonstrate that our method achieves state-of-the-art performance with high efficiency. Our code and model checkpoints will be available upon acceptance.
2502.10038
POI-Enhancer: An LLM-based Semantic Enhancement Framework for POI Representation Learning
cs.AI
POI representation learning plays a crucial role in handling tasks related to user mobility data. Recent studies have shown that enriching POI representations with multimodal information can significantly enhance their task performance. Previously, the textual information incorporated into POI representations typically involved only POI categories or check-in content, leading to relatively weak textual features in existing methods. In contrast, large language models (LLMs) trained on extensive text data have been found to possess rich textual knowledge. However leveraging such knowledge to enhance POI representation learning presents two key challenges: first, how to extract POI-related knowledge from LLMs effectively, and second, how to integrate the extracted information to enhance POI representations. To address these challenges, we propose POI-Enhancer, a portable framework that leverages LLMs to improve POI representations produced by classic POI learning models. We first design three specialized prompts to extract semantic information from LLMs efficiently. Then, the Dual Feature Alignment module enhances the quality of the extracted information, while the Semantic Feature Fusion module preserves its integrity. The Cross Attention Fusion module then fully adaptively integrates such high-quality information into POI representations and Multi-View Contrastive Learning further injects human-understandable semantic information into these representations. Extensive experiments on three real-world datasets demonstrate the effectiveness of our framework, showing significant improvements across all baseline representations.
2502.10040
Diffusion Trajectory-guided Policy for Long-horizon Robot Manipulation
cs.RO
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.
2502.10042
Scaling Law Tradeoff Between Throughput and Sensing Distance in Large ISAC Networks
cs.IT math.IT
In this paper, we investigate the fundamental tradeoff between communication and sensing performance of \emph{ad hoc} integrated sensing and communication (ISAC) wireless networks. Specifically, we consider that $n$ nodes are randomly located in an extended network with area $n$ and transmit ISAC signals. Under the pure path loss channel gain model and the condition that the transmission power scales according to the communication distance, we fully characterize the optimal scaling law tradeoff between throughput and sensing distance by proposing an achievable scheme and proving its converse. Our results can be interpreted as follows: by reducing the throughput by a factor of a function of $n$, the sensing range order improves according to the same function of $n$, raised to the power of the ratio between the path loss factors in communication and sensing. We prove that the same result also holds true for ISAC networks with random fading, despite the uncertainty on the connectivity and power level created by random fading. In addition, we show that the scaling law tradeoff cannot be improved by allowing the transmission power and communication distance to scale freely. To the best of our knowledge, this is the first work formally formulating and characterizing the communication and sensing performance scaling law tradeoff of \emph{ad hoc} ISAC networks.
2502.10044
Unsupervised Entity Alignment Based on Personalized Discriminative Rooted Tree
cs.AI
Entity Alignment (EA) is to link potential equivalent entities across different knowledge graphs (KGs). Most existing EA methods are supervised as they require the supervision of seed alignments, i.e., manually specified aligned entity pairs. Very recently, several EA studies have made some attempts to get rid of seed alignments. Despite achieving preliminary progress, they still suffer two limitations: (1) The entity embeddings produced by their GNN-like encoders lack personalization since some of the aggregation subpaths are shared between different entities. (2) They cannot fully alleviate the distribution distortion issue between candidate KGs due to the absence of the supervised signal. In this work, we propose a novel unsupervised entity alignment approach called UNEA to address the above two issues. First, we parametrically sample a tree neighborhood rooted at each entity, and accordingly develop a tree attention aggregation mechanism to extract a personalized embedding for each entity. Second, we introduce an auxiliary task of maximizing the mutual information between the input and the output of the KG encoder, to regularize the model and prevent the distribution distortion. Extensive experiments show that our UNEA achieves a new state-of-the-art for the unsupervised EA task, and can even outperform many existing supervised EA baselines.
2502.10046
ViRAC: A Vision-Reasoning Agent Head Movement Control Framework in Arbitrary Virtual Environments
cs.GR cs.CV
Creating lifelike virtual agents capable of interacting with their environments is a longstanding goal in computer graphics. This paper addresses the challenge of generating natural head rotations, a critical aspect of believable agent behavior for visual information gathering and dynamic responses to environmental cues. Although earlier methods have made significant strides, many rely on data-driven or saliency-based approaches, which often underperform in diverse settings and fail to capture deeper cognitive factors such as risk assessment, information seeking, and contextual prioritization. Consequently, generated behaviors can appear rigid or overlook critical scene elements, thereby diminishing the sense of realism. In this paper, we propose \textbf{ViRAC}, a \textbf{Vi}sion-\textbf{R}easoning \textbf{A}gent Head Movement \textbf{C}ontrol framework, which exploits the common-sense knowledge and reasoning capabilities of large-scale models, including Vision-Language Models (VLMs) and Large-Language Models (LLMs). Rather than explicitly modeling every cognitive mechanism, ViRAC leverages the biases and patterns internalized by these models from extensive training, thus emulating human-like perceptual processes without hand-tuned heuristics. Experimental results in multiple scenarios reveal that ViRAC produces more natural and context-aware head rotations than recent state-of-the-art techniques. Quantitative evaluations show a closer alignment with real human head-movement data, while user studies confirm improved realism and cognitive plausibility.
2502.10047
Janus: Collaborative Vision Transformer Under Dynamic Network Environment
cs.DC cs.AI
Vision Transformers (ViTs) have outperformed traditional Convolutional Neural Network architectures and achieved state-of-the-art results in various computer vision tasks. Since ViTs are computationally expensive, the models either have to be pruned to run on resource-limited edge devices only or have to be executed on remote cloud servers after receiving the raw data transmitted over fluctuating networks. The resulting degraded performance or high latency all hinder their widespread applications. In this paper, we present Janus, the first framework for low-latency cloud-device collaborative Vision Transformer inference over dynamic networks. Janus overcomes the intrinsic model limitations of ViTs and realizes collaboratively executing ViT models on both cloud and edge devices, achieving low latency, high accuracy, and low communication overhead. Specifically, Janus judiciously combines token pruning techniques with a carefully designed fine-to-coarse model splitting policy and non-static mixed pruning policy. It attains a balance between accuracy and latency by dynamically selecting the optimal pruning level and split point. Experimental results across various tasks demonstrate that Janus enhances throughput by up to 5.15 times and reduces latency violation ratios by up to 98.7% when compared with baseline approaches under various network environments.
2502.10050
A Survey on LLM-powered Agents for Recommender Systems
cs.IR cs.AI
Recommender systems are essential components of many online platforms, yet traditional approaches still struggle with understanding complex user preferences and providing explainable recommendations. The emergence of Large Language Model (LLM)-powered agents offers a promising approach by enabling natural language interactions and interpretable reasoning, potentially transforming research in recommender systems. This survey provides a systematic review of the emerging applications of LLM-powered agents in recommender systems. We identify and analyze three key paradigms in current research: (1) Recommender-oriented approaches, which leverage intelligent agents to enhance the fundamental recommendation mechanisms; (2) Interaction-oriented approaches, which facilitate dynamic user engagement through natural dialogue and interpretable suggestions; and (3) Simulation-oriented approaches, which employ multi-agent frameworks to model complex user-item interactions and system dynamics. Beyond paradigm categorization, we analyze the architectural foundations of LLM-powered recommendation agents, examining their essential components: profile construction, memory management, strategic planning, and action execution. Our investigation extends to a comprehensive analysis of benchmark datasets and evaluation frameworks in this domain. This systematic examination not only illuminates the current state of LLM-powered agent recommender systems but also charts critical challenges and promising research directions in this transformative field.
2502.10051
ORI: O Routing Intelligence
cs.CL
Single large language models (LLMs) often fall short when faced with the ever-growing range of tasks, making a single-model approach insufficient. We address this challenge by proposing ORI (O Routing Intelligence), a dynamic framework that leverages a set of LLMs. By intelligently routing incoming queries to the most suitable model, ORI not only improves task-specific accuracy, but also maintains efficiency. Comprehensive evaluations across diverse benchmarks demonstrate consistent accuracy gains while controlling computational overhead. By intelligently routing queries, ORI outperforms the strongest individual models by up to 2.7 points on MMLU and 1.8 points on MuSR, ties the top performance on ARC, and on BBH. These results underscore the benefits of a multi-model strategy and demonstrate how ORI's adaptive architecture can more effectively handle diverse tasks, offering a scalable, high-performance solution for a system of multiple large language models.
2502.10054
Towards Polyp Counting In Full-Procedure Colonoscopy Videos
cs.CV
Automated colonoscopy reporting holds great potential for enhancing quality control and improving cost-effectiveness of colonoscopy procedures. A major challenge lies in the automated identification, tracking, and re-association (ReID) of polyps tracklets across full-procedure colonoscopy videos. This is essential for precise polyp counting and enables automated computation of key quality metrics, such as Adenoma Detection Rate (ADR) and Polyps Per Colonoscopy (PPC). However, polyp ReID is challenging due to variations in polyp appearance, frequent disappearance from the field of view, and occlusions. In this work, we leverage the REAL-Colon dataset, the first open-access dataset providing full-procedure videos, to define tasks, data splits and metrics for the problem of automatically count polyps in full-procedure videos, establishing an open-access framework. We re-implement previously proposed SimCLR-based methods for learning representations of polyp tracklets, both single-frame and multi-view, and adapt them to the polyp counting task. We then propose an Affinity Propagation-based clustering method to further improve ReID based on these learned representations, ultimately enhancing polyp counting. Our approach achieves state-of-the-art performance, with a polyp fragmentation rate of 6.30 and a false positive rate (FPR) below 5% on the REAL-Colon dataset. We release code at https://github.com/lparolari/towards-polyp-counting.
2502.10057
A Generalized Modeling Approach to Liquid-driven Ballooning Membranes
cs.RO
Soft robotics is advancing the use of flexible materials for adaptable robotic systems. Membrane-actuated soft robots address the limitations of traditional soft robots by using pressurized, extensible membranes to achieve stable, large deformations, yet control and state estimation remain challenging due to their complex deformation dynamics. This paper presents a novel modeling approach for liquid-driven ballooning membranes, employing an ellipsoid approximation to model shape and stretch under planar deformation. Relying solely on intrinsic feedback from pressure data and controlled liquid volume, this approach enables accurate membrane state estimation. We demonstrate the effectiveness of the proposed model for ballooning membrane-based actuators by experimental validation, obtaining the indentation depth error of $RMSE_{h_2}=0.80\;$mm, which is $23\%$ of the indentation range and $6.67\%$ of the unindented actuator height range. For the force estimation, the error range is obtained to be $RMSE_{F}=0.15\;$N which is $10\%$ of the measured force range.
2502.10058
MTLM: an Innovative Language Model Training Paradigm for ASR
cs.CL eess.AS
Pre-training Transformer-based language models (LMs) on a large amount of text has proven crucial for improving automatic speech recognition (ASR) performance. Generally, traditional LMs are unidirectional and unable to access the context on the right. This paper proposes a method for training LMs that enable traditional unidirectional LMs to fully utilize left and right contexts. Compared with the unidirectional LMs, our LM facilitates ASR to transcribe hypotheses more consistently and in a more semantically unambiguous way, as it incorporates richer contextual representations. Finally, our experimental results on the LibriSpeech corpus demonstrate that our model outperforms traditional unidirectional LMs, whether n-best rescoring or shallow fusion is used as the decoding algorithm.
2502.10059
RealCam-I2V: Real-World Image-to-Video Generation with Interactive Complex Camera Control
cs.CV
Recent advancements in camera-trajectory-guided image-to-video generation offer higher precision and better support for complex camera control compared to text-based approaches. However, they also introduce significant usability challenges, as users often struggle to provide precise camera parameters when working with arbitrary real-world images without knowledge of their depth nor scene scale. To address these real-world application issues, we propose RealCam-I2V, a novel diffusion-based video generation framework that integrates monocular metric depth estimation to establish 3D scene reconstruction in a preprocessing step. During training, the reconstructed 3D scene enables scaling camera parameters from relative to absolute values, ensuring compatibility and scale consistency across diverse real-world images. In inference, RealCam-I2V offers an intuitive interface where users can precisely draw camera trajectories by dragging within the 3D scene. To further enhance precise camera control and scene consistency, we propose scene-constrained noise shaping, which shapes high-level noise and also allows the framework to maintain dynamic, coherent video generation in lower noise stages. RealCam-I2V achieves significant improvements in controllability and video quality on the RealEstate10K and out-of-domain images. We further enables applications like camera-controlled looping video generation and generative frame interpolation. We will release our absolute-scale annotation, codes, and all checkpoints. Please see dynamic results in https://zgctroy.github.io/RealCam-I2V.
2502.10060
DiSciPLE: Learning Interpretable Programs for Scientific Visual Discovery
cs.CV cs.LG
Visual data is used in numerous different scientific workflows ranging from remote sensing to ecology. As the amount of observation data increases, the challenge is not just to make accurate predictions but also to understand the underlying mechanisms for those predictions. Good interpretation is important in scientific workflows, as it allows for better decision-making by providing insights into the data. This paper introduces an automatic way of obtaining such interpretable-by-design models, by learning programs that interleave neural networks. We propose DiSciPLE (Discovering Scientific Programs using LLMs and Evolution) an evolutionary algorithm that leverages common sense and prior knowledge of large language models (LLMs) to create Python programs explaining visual data. Additionally, we propose two improvements: a program critic and a program simplifier to improve our method further to synthesize good programs. On three different real-world problems, DiSciPLE learns state-of-the-art programs on novel tasks with no prior literature. For example, we can learn programs with 35% lower error than the closest non-interpretable baseline for population density estimation.
2502.10061
Annotating Compositionality Scores for Irish Noun Compounds is Hard Work
cs.CL
Noun compounds constitute a challenging construction for NLP applications, given their variability in idiomaticity and interpretation. In this paper, we present an analysis of compound nouns identified in Irish text of varied domains by expert annotators, focusing on compositionality as a key feature, but also domain specificity, as well as familiarity and confidence of the annotator giving the ratings. Our findings and the discussion that ensued contributes towards a greater understanding of how these constructions appear in Irish language, and how they might be treated separately from English noun compounds.
2502.10062
Adaptive Bi-Level Multi-Robot Task Allocation and Learning under Uncertainty with Temporal Logic Constraints
cs.RO cs.AI cs.FL
This work addresses the problem of multi-robot coordination under unknown robot transition models, ensuring that tasks specified by Time Window Temporal Logic are satisfied with user-defined probability thresholds. We present a bi-level framework that integrates (i) high-level task allocation, where tasks are assigned based on the robots' estimated task completion probabilities and expected rewards, and (ii) low-level distributed policy learning and execution, where robots independently optimize auxiliary rewards while fulfilling their assigned tasks. To handle uncertainty in robot dynamics, our approach leverages real-time task execution data to iteratively refine expected task completion probabilities and rewards, enabling adaptive task allocation without explicit robot transition models. We theoretically validate the proposed algorithm, demonstrating that the task assignments meet the desired probability thresholds with high confidence. Finally, we demonstrate the effectiveness of our framework through comprehensive simulations.
2502.10063
Strassen Multisystolic Array Hardware Architectures
cs.AR cs.AI cs.PF
While Strassen's matrix multiplication algorithm reduces the complexity of naive matrix multiplication, general-purpose hardware is not suitable for achieving the algorithm's promised theoretical speedups. This leaves the question of if it could be better exploited in custom hardware architectures designed specifically for executing the algorithm. However, there is limited prior work on this and it is not immediately clear how to derive such architectures or if they can ultimately lead to real improvements. We bridge this gap, presenting and evaluating new systolic array architectures that efficiently translate the theoretical complexity reductions of Strassen's algorithm directly into hardware resource savings. Furthermore, the architectures are multisystolic array designs that can multiply smaller matrices with higher utilization than single-systolic array designs. The proposed designs implemented on FPGA reduce DSP requirements by a factor of $1.14^r$ for $r$ implemented Strassen recursion levels, and otherwise require overall similar soft logic resources when instantiated to support matrix sizes down to 32x32 and 24x24 at 1-2 levels of Strassen recursion, respectively. We evaluate the proposed designs both in isolation and in an end-to-end machine learning accelerator compared to baseline designs and prior works, achieving state-of-the-art performance.
2502.10064
Hands-off Image Editing: Language-guided Editing without any Task-specific Labeling, Masking or even Training
cs.CL cs.CV
Instruction-guided image editing consists in taking an image and an instruction and deliverring that image altered according to that instruction. State-of-the-art approaches to this task suffer from the typical scaling up and domain adaptation hindrances related to supervision as they eventually resort to some kind of task-specific labelling, masking or training. We propose a novel approach that does without any such task-specific supervision and offers thus a better potential for improvement. Its assessment demonstrates that it is highly effective, achieving very competitive performance.
2502.10070
Topological Neural Networks over the Air
cs.IT cs.LG math.IT
Topological neural networks (TNNs) are information processing architectures that model representations from data lying over topological spaces (e.g., simplicial or cell complexes) and allow for decentralized implementation through localized communications over different neighborhoods. Existing TNN architectures have not yet been considered in realistic communication scenarios, where channel effects typically introduce disturbances such as fading and noise. This paper aims to propose a novel TNN design, operating on regular cell complexes, that performs over-the-air computation, incorporating the wireless communication model into its architecture. Specifically, during training and inference, the proposed method considers channel impairments such as fading and noise in the topological convolutional filtering operation, which takes place over different signal orders and neighborhoods. Numerical results illustrate the architecture's robustness to channel impairments during testing and the superior performance with respect to existing architectures, which are either communication-agnostic or graph-based.
2502.10072
LifeSaver: Predictive Load Limit Estimation for Transport Vehicles in Hilly Areas
eess.SY cs.SY
The transportation of essential goods in mountainous regions faces severe logistical challenges and frequent disruptions. To mitigate these difficulties, transport companies often overload trucks, which, though cost-saving, significantly heightens the risk of accidents and mechanical failures. This paper presents the development of a device that detects overloaded and insecurely fastened loads on trucks and commercial vehicles. Using advanced load sensors, the device offers real-time monitoring of cargo weight distribution, alerting drivers and authorities to unsafe conditions. The initial prototype utilised two basic load cells and an Arduino microcontroller. The second version was enhanced with four load cells and extended sensors. This version was tested by placing an electric golf cart onto the prototype. Various loads were then added to the cart in different orientations to assess whether the system could accurately detect improper or excessive load conditions.
2502.10076
Classification of Temporal Graphs using Persistent Homology
cs.LG cs.CG math.AT
Temporal graphs effectively model dynamic systems by representing interactions as timestamped edges. However, analytical tools for temporal graphs are limited compared to static graphs. We propose a novel method for analyzing temporal graphs using Persistent Homology. Our approach leverages $\delta$-temporal motifs (recurrent subgraphs) to capture temporal dynamics %without aggregation . By evolving these motifs, we define the \textit{average filtration} and compute PH on the associated clique complex. This method captures both local and global temporal structures and is stable with respect to reference models. We demonstrate the applicability of our approach to the temporal graph classification task. Experiments verify the effectiveness of our approach, achieving over 92\% accuracy, with some cases reaching 100\%. Unlike existing methods that require node classes, our approach is node class free, offering flexibility for a wide range of temporal graph analysis.
2502.10077
Towards Empowerment Gain through Causal Structure Learning in Model-Based RL
cs.AI cs.LG
In Model-Based Reinforcement Learning (MBRL), incorporating causal structures into dynamics models provides agents with a structured understanding of the environments, enabling efficient decision. Empowerment as an intrinsic motivation enhances the ability of agents to actively control their environments by maximizing the mutual information between future states and actions. We posit that empowerment coupled with causal understanding can improve controllability, while enhanced empowerment gain can further facilitate causal reasoning in MBRL. To improve learning efficiency and controllability, we propose a novel framework, Empowerment through Causal Learning (ECL), where an agent with the awareness of causal dynamics models achieves empowerment-driven exploration and optimizes its causal structure for task learning. Specifically, ECL operates by first training a causal dynamics model of the environment based on collected data. We then maximize empowerment under the causal structure for exploration, simultaneously using data gathered through exploration to update causal dynamics model to be more controllable than dense dynamics model without causal structure. In downstream task learning, an intrinsic curiosity reward is included to balance the causality, mitigating overfitting. Importantly, ECL is method-agnostic and is capable of integrating various causal discovery methods. We evaluate ECL combined with 3 causal discovery methods across 6 environments including pixel-based tasks, demonstrating its superior performance compared to other causal MBRL methods, in terms of causal discovery, sample efficiency, and asymptotic performance.
2502.10080
Coordinated control of multiple autonomous surface vehicles: challenges and advances -- a systematic review
cs.RO cs.SY eess.SY
The increasing use and implementation of Autonomous Surface Vessels (ASVs) for various activities in maritime environments is expected to drive a rise in developments and research on their control. Particularly, the coordination of multiple ASVs presents novel challenges and opportunities, requiring interdisciplinary research efforts at the intersection of robotics, control theory, communication systems, and marine sciences. The wide variety of missions or objectives for which these vessels can be collectively used allows for the application and combination of different control techniques. This includes the exploration of machine learning to consider aspects previously deemed infeasible. This review provides a comprehensive exploration of coordinated ASV control while addressing critical gaps left by previous reviews. Unlike previous works, we adopt a systematic approach to ensure integrity and minimize bias in article selection. We delve into the complex world of sub-actuated ASVs with a focus on customized control strategies and the integration of machine learning techniques for increased autonomy. By synthesizing recent advances and identifying emerging trends, we offer insights that drive this field forward, providing both a comprehensive overview of state-of-the-art techniques and guidance for future research efforts.
2502.10089
A Hybrid Edge Classifier: Combining TinyML-Optimised CNN with RRAM-CMOS ACAM for Energy-Efficient Inference
cs.LG cs.AI cs.AR
In recent years, the development of smart edge computing systems to process information locally is on the rise. Many near-sensor machine learning (ML) approaches have been implemented to introduce accurate and energy efficient template matching operations in resource-constrained edge sensing systems, such as wearables. To introduce novel solutions that can be viable for extreme edge cases, hybrid solutions combining conventional and emerging technologies have started to be proposed. Deep Neural Networks (DNN) optimised for edge application alongside new approaches of computing (both device and architecture -wise) could be a strong candidate in implementing edge ML solutions that aim at competitive accuracy classification while using a fraction of the power of conventional ML solutions. In this work, we are proposing a hybrid software-hardware edge classifier aimed at the extreme edge near-sensor systems. The classifier consists of two parts: (i) an optimised digital tinyML network, working as a front-end feature extractor, and (ii) a back-end RRAM-CMOS analogue content addressable memory (ACAM), working as a final stage template matching system. The combined hybrid system exhibits a competitive trade-off in accuracy versus energy metric with $E_{front-end}$ = $96.23 nJ$ and $E_{back-end}$ = $1.45 nJ$ for each classification operation compared with 78.06$\mu$J for the original teacher model, representing a 792-fold reduction, making it a viable solution for extreme edge applications.
2502.10090
Manual2Skill: Learning to Read Manuals and Acquire Robotic Skills for Furniture Assembly Using Vision-Language Models
cs.RO cs.AI
Humans possess an extraordinary ability to understand and execute complex manipulation tasks by interpreting abstract instruction manuals. For robots, however, this capability remains a substantial challenge, as they cannot interpret abstract instructions and translate them into executable actions. In this paper, we present Manual2Skill, a novel framework that enables robots to perform complex assembly tasks guided by high-level manual instructions. Our approach leverages a Vision-Language Model (VLM) to extract structured information from instructional images and then uses this information to construct hierarchical assembly graphs. These graphs represent parts, subassemblies, and the relationships between them. To facilitate task execution, a pose estimation model predicts the relative 6D poses of components at each assembly step. At the same time, a motion planning module generates actionable sequences for real-world robotic implementation. We demonstrate the effectiveness of Manual2Skill by successfully assembling several real-world IKEA furniture items. This application highlights its ability to manage long-horizon manipulation tasks with both efficiency and precision, significantly enhancing the practicality of robot learning from instruction manuals. This work marks a step forward in advancing robotic systems capable of understanding and executing complex manipulation tasks in a manner akin to human capabilities.
2502.10091
ELAA-ISAC: Environmental Mapping Utilizing the LoS State of Communication Channel
cs.IT eess.SP math.IT
In this paper, a novel environmental mapping method is proposed to outline the indoor layout utilizing the line-of-sight (LoS) state information of extremely large aperture array (ELAA) channels. It leverages the spatial resolution provided by ELAA and the mobile terminal (MT)'s mobility to infer the presence and location of obstacles in the environment. The LoS state estimation is formulated as a binary hypothesis testing problem, and the optimal decision rule is derived based on the likelihood ratio test. Subsequently, the theoretical error probability of LoS estimation is derived, showing close alignment with simulation results. Then, an environmental mapping method is proposed, which progressively outlines the layout by combining LoS state information from multiple MT locations. It is demonstrated that the proposed method can accurately outline the environment layout, with the mapping accuracy improving as the number of service-antennas and MT locations increases. This paper also investigates the impact of channel estimation error and non-LoS (NLoS) components on the quality of environmental mapping. The proposed method exhibits particularly promising performance in LoS dominated wireless environments characterized by high Rician K-factor. Specifically, it achieves an average intersection over union (IoU) exceeding 80% when utilizing 256 service antennas and 18 MT locations.
2502.10092
A novel approach to data generation in generative model
cs.LG cs.AI
Variational Autoencoders (VAEs) and other generative models are widely employed in artificial intelligence to synthesize new data. However, current approaches rely on Euclidean geometric assumptions and statistical approximations that fail to capture the structured and emergent nature of data generation. This paper introduces the Convergent Fusion Paradigm (CFP) theory, a novel geometric framework that redefines data generation by integrating dimensional expansion accompanied by qualitative transformation. By modifying the latent space geometry to interact with emergent high-dimensional structures, CFP theory addresses key challenges such as identifiability issues and unintended artifacts like hallucinations in Large Language Models (LLMs). CFP theory is based on two key conceptual hypotheses that redefine how generative models structure relationships between data and algorithms. Through the lens of CFP theory, we critically examine existing metric-learning approaches. CFP theory advances this perspective by introducing time-reversed metric embeddings and structural convergence mechanisms, leading to a novel geometric approach that better accounts for data generation as a structured epistemic process. Beyond its computational implications, CFP theory provides philosophical insights into the ontological underpinnings of data generation. By offering a systematic framework for high-dimensional learning dynamics, CFP theory contributes to establishing a theoretical foundation for understanding the data-relationship structures in AI. Finally, future research in CFP theory will be led to its implications for fully realizing qualitative transformations, introducing the potential of Hilbert space in generative modeling.
2502.10095
Representation Learning on Out of Distribution in Tabular Data
cs.LG
The open-world assumption in model development suggests that a model might lack sufficient information to adequately handle data that is entirely distinct or out of distribution (OOD). While deep learning methods have shown promising results in handling OOD data through generalization techniques, they often require specialized hardware that may not be accessible to all users. We present TCL, a lightweight yet effective solution that operates efficiently on standard CPU hardware. Our approach adapts contrastive learning principles specifically for tabular data structures, incorporating full matrix augmentation and simplified loss calculation. Through comprehensive experiments across 10 diverse datasets, we demonstrate that TCL outperforms existing models, including FT-Transformer and ResNet, particularly in classification tasks, while maintaining competitive performance in regression problems. TCL achieves these results with significantly reduced computational requirements, making it accessible to users with limited hardware capabilities. This study also provides practical guidance for detecting and evaluating OOD data through straightforward experiments and visualizations. Our findings show that TCL offers a promising balance between performance and efficiency in handling OOD prediction tasks, which is particularly beneficial for general machine learning practitioners working with computational constraints.
2502.10097
Causal Information Prioritization for Efficient Reinforcement Learning
cs.AI cs.LG
Current Reinforcement Learning (RL) methods often suffer from sample-inefficiency, resulting from blind exploration strategies that neglect causal relationships among states, actions, and rewards. Although recent causal approaches aim to address this problem, they lack grounded modeling of reward-guided causal understanding of states and actions for goal-orientation, thus impairing learning efficiency. To tackle this issue, we propose a novel method named Causal Information Prioritization (CIP) that improves sample efficiency by leveraging factored MDPs to infer causal relationships between different dimensions of states and actions with respect to rewards, enabling the prioritization of causal information. Specifically, CIP identifies and leverages causal relationships between states and rewards to execute counterfactual data augmentation to prioritize high-impact state features under the causal understanding of the environments. Moreover, CIP integrates a causality-aware empowerment learning objective, which significantly enhances the agent's execution of reward-guided actions for more efficient exploration in complex environments. To fully assess the effectiveness of CIP, we conduct extensive experiments across 39 tasks in 5 diverse continuous control environments, encompassing both locomotion and manipulation skills learning with pixel-based and sparse reward settings. Experimental results demonstrate that CIP consistently outperforms existing RL methods across a wide range of scenarios.
2502.10100
Statistical data analysis for Tourism in Poland in R Programming Environment
math.NA cs.CE cs.ET cs.NA cs.PL
This study utilises the R programming language for statistical data analysis to understand Tourism dynamics in Poland. It focuses on methods for data visualisation, multivariate statistics, and hypothesis testing. To investigate the expenditure behavior of tourist, spending patterns, correlations, and associations among variables were analysed in the dataset. The results revealed a significant relationship between accommodation type and the purpose of trip, showing that the purpose of a trip impacts the selection of accommodation. A strong correlation was observed between organizer expenditure and private expenditure, indicating that individual spending are more when the spending on organizing the trip are higher. However, no significant difference was observed in total expenditure across different accommodation types and purpose of the trip revealing that travelers tend to spend similar amounts regardless of their reason for travel or choice of accommodation. Although significant relationships were observed among certain variables, ANOVA could not be applied because the dataset was not able to hold on the normality assumption. In future, the dataset can be explored further to find more meaningful insights. The developed code is available on GitHub: https://github.com/SaadAhmedJamal/DataAnalysis RProgEnv.
2502.10106
Data-Adaptive Low-Rank Sparse Subspace Clustering
cs.LG
Low-rank sparse subspace clustering (LRSSC) algorithms built on self-expressive model effectively capture both the global and local structure of the data. However, existing solutions, primarily based on proximal operators associated with Sp/Lp , p e {0, 1/2, 2/3, 1}, norms are not data-adaptive. In this work, we propose an LRSSC algorithm incorporating a data-adaptive surrogate for the S0/L0 quasi-norm. We provide a numerical solution for the corresponding proximal operator in cases where an analytical expression is unavailable. The proposed LRSSC algorithm is formulated within the proximal mapping framework, and we present theoretical proof of its global convergence toward a stationary point. We evaluate the performance of the proposed method on three well known datasets, comparing it against LRSSC algorithms constrained by Sp/Lp, p e {0, 1/2, 2/3, 1}, norms.
2502.10108
NeuroXVocal: Detection and Explanation of Alzheimer's Disease through Non-invasive Analysis of Picture-prompted Speech
cs.LG q-bio.NC
The early diagnosis of Alzheimer's Disease (AD) through non invasive methods remains a significant healthcare challenge. We present NeuroXVocal, a novel dual-component system that not only classifies but also explains potential AD cases through speech analysis. The classification component (Neuro) processes three distinct data streams: acoustic features capturing speech patterns and voice characteristics, textual features extracted from speech transcriptions, and precomputed embeddings representing linguistic patterns. These streams are fused through a custom transformer-based architecture that enables robust cross-modal interactions. The explainability component (XVocal) implements a Retrieval-Augmented Generation (RAG) approach, leveraging Large Language Models combined with a domain-specific knowledge base of AD research literature. This architecture enables XVocal to retrieve relevant clinical studies and research findings to generate evidence-based context-sensitive explanations of the acoustic and linguistic markers identified in patient speech. Using the IS2021 ADReSSo Challenge benchmark dataset, our system achieved state-of-the-art performance with 95.77% accuracy in AD classification, significantly outperforming previous approaches. The explainability component was qualitatively evaluated using a structured questionnaire completed by medical professionals, validating its clinical relevance. NeuroXVocal's unique combination of high-accuracy classification and interpretable, literature-grounded explanations demonstrates its potential as a practical tool for supporting clinical AD diagnosis.
2502.10111
COMBINEX: A Unified Counterfactual Explainer for Graph Neural Networks via Node Feature and Structural Perturbations
cs.LG
Counterfactual explanations have emerged as a powerful tool to unveil the opaque decision-making processes of graph neural networks (GNNs). However, existing techniques primarily focus on edge modifications, often overlooking the crucial role of node feature perturbations in shaping model predictions. To address this limitation, we propose COMBINEX, a novel GNN explainer that generates counterfactual explanations for both node and graph classification tasks. Unlike prior methods, which treat structural and feature-based changes independently, COMBINEX optimally balances modifications to edges and node features by jointly optimizing these perturbations. This unified approach ensures minimal yet effective changes required to flip a model's prediction, resulting in realistic and interpretable counterfactuals. Additionally, COMBINEX seamlessly handles both continuous and discrete node features, enhancing its versatility across diverse datasets and GNN architectures. Extensive experiments on real-world datasets and various GNN architectures demonstrate the effectiveness and robustness of our approach over existing baselines.
2502.10112
Accelerometry-based Energy Expenditure Estimation During Activities of Daily Living: A Comparison Among Different Accelerometer Compositions
cs.LG
Physical activity energy expenditure (PAEE) can be measured from breath-by-breath respiratory data, which can serve as a reference. Alternatively, PAEE can be predicted from the body movements, which can be measured and estimated with accelerometers. The body center of mass (COM) acceleration reflects the movements of the whole body and thus serves as a good predictor for PAEE. However, the wrist has also become a popular location due to recent advancements in wrist-worn devices. Therefore, in this work, using the respiratory data measured by COSMED K5 as the reference, we evaluated and compared the performances of COM-based settings and wrist-based settings. The COM-based settings include two different accelerometer compositions, using only the pelvis accelerometer (pelvis-acc) and the pelvis accelerometer with two accelerometers from two thighs (3-acc). The wrist-based settings include using only the left wrist accelerometer (l-wrist-acc) and only the right wrist accelerometer (r-wrist-acc). We implemented two existing PAEE estimation methods on our collected dataset, where 9 participants performed activities of daily living while wearing 5 accelerometers (i.e., pelvis, two thighs, and two wrists). These two methods include a linear regression (LR) model and a CNN-LSTM model. Both models yielded the best results with the COM-based 3-acc setting (LR: $R^2$ = 0.41, CNN-LSTM: $R^2$ = 0.53). No significant difference was found between the 3-acc and pelvis-acc settings (p-value = 0.278). For both models, neither the l-wrist-acc nor the r-wrist-acc settings demonstrated predictive power on PAEE with $R^2$ values close to 0, significantly outperformed by the two COM-based settings (p-values $<$ 0.05). No significant difference was found between the two wrists (p-value = 0.329).
2502.10113
Strain-Induced Optical and Molecular Transformations in PET Films for Organic Electronic Applications
physics.app-ph cs.SY eess.SY physics.optics
Poly(ethylene terephthalate) (PET) films are widely used in flexible electronics and optoelectronics, where their mechanical durability and optical performance under strain are essential for device reliability. This study investigates the impact of applied mechanical strain on the optical and molecular properties of PET at room temperature,using UV-Vis absorption and Raman spectroscopy. The work explores how varying strain levels, from 0% (unstretched) to 30%, affect the transparency, vibrational modes, and molecular reorganization within PET films. UV-Vis absorbance measurements reveal that strain induces significant changes in the light transmission properties of PET, particularly in the visible range, and increases absorption in the UVA and visible region by up to 100%. Raman spectra indicate that strain levels higher than 5% lead to irreversible shifts of vibrational lines, accompanied by an increase of their full width at half maximum (FWHM), suggesting molecular reorientation and crystallinity changes. The phonon mode coupled with C-O stretching [O-CH2] shows the strongest response to applied mechanical stress. This study provides a comprehensive understanding of strain-induced optical and structural alterations in PET, with implications for improving the mechanical and optical performance of PET-based devices in strainsensitive applications, such as organic solar cells (OSCs), organic light-emitting diodes (OLEDs), and flexible sensors.
2502.10118
Image Embedding Sampling Method for Diverse Captioning
cs.CV cs.AI
Image Captioning for state-of-the-art VLMs has significantly improved over time; however, this comes at the cost of increased computational complexity, making them less accessible for resource-constrained applications such as mobile devices and assistive technologies. Alternatively, smaller VLMs prioritize high-level scene descriptions, overlooking finer details that contribute to a richer understanding of an image. In this paper, we introduce a training-free framework that enhances caption diversity and informativeness by explicitly attending to distinct image regions using a comparably small VLM, BLIP, as the backbone. Our approach leverages structured segmentation to produce hierarchical representations that capture both global and localized semantics. Without requiring additional model training, we demonstrate that our method allows smaller VLMs to achieve performance comparable to larger models in terms of image-caption alignment, semantic integrity, and diversity. We evaluate our framework on MSCOCO, Flickr30k, and Nocaps test datasets, achieving a Div-2 score of 0.735, 0.750, and 0.748 for each dataset respectively, while maintaining strong image-caption relevancy and semantic integrity with the human-annotated captions.
2502.10119
SeWA: Selective Weight Average via Probabilistic Masking
cs.LG
Weight averaging has become a standard technique for enhancing model performance. However, methods such as Stochastic Weight Averaging (SWA) and Latest Weight Averaging (LAWA) often require manually designed procedures to sample from the training trajectory, and the results depend heavily on hyperparameter tuning. To minimize human effort, this paper proposes a simple yet efficient algorithm called Selective Weight Averaging (SeWA), which adaptively selects checkpoints during the final stages of training for averaging. Based on SeWA, we show that only a few points are needed to achieve better generalization and faster convergence. Theoretically, solving the discrete subset selection problem is inherently challenging. To address this, we transform it into a continuous probabilistic optimization framework and employ the Gumbel-Softmax estimator to learn the non-differentiable mask for each checkpoint. Further, we theoretically derive the SeWA's stability-based generalization bounds, which are sharper than that of SGD under both convex and non-convex assumptions. Finally, solid extended experiments in various domains, including behavior cloning, image classification, and text classification, further validate the effectiveness of our approach.
2502.10120
Compress image to patches for Vision Transformer
cs.CV
The Vision Transformer (ViT) has made significant strides in the field of computer vision. However, as the depth of the model and the resolution of the input images increase, the computational cost associated with training and running ViT models has surged dramatically. This paper proposes a hybrid model based on CNN and Vision Transformer, named CI2P-ViT. The model incorporates a module called CI2P, which utilizes the CompressAI encoder to compress images and subsequently generates a sequence of patches through a series of convolutions. CI2P can replace the Patch Embedding component in the ViT model, enabling seamless integration into existing ViT models. Compared to ViT-B/16, CI2P-ViT has the number of patches input to the self-attention layer reduced to a quarter of the original. This design not only significantly reduces the computational cost of the ViT model but also effectively enhances the model's accuracy by introducing the inductive bias properties of CNN. The ViT model's precision is markedly enhanced. When trained from the ground up on the Animals-10 dataset, CI2P-ViT achieved an accuracy rate of 92.37%, representing a 3.3% improvement over the ViT-B/16 baseline. Additionally, the model's computational operations, measured in floating-point operations per second (FLOPs), were diminished by 63.35%, and it exhibited a 2-fold increase in training velocity on identical hardware configurations.
2502.10122
Modern Hopfield Networks with Continuous-Time Memories
cs.LG
Recent research has established a connection between modern Hopfield networks (HNs) and transformer attention heads, with guarantees of exponential storage capacity. However, these models still face challenges scaling storage efficiently. Inspired by psychological theories of continuous neural resource allocation in working memory, we propose an approach that compresses large discrete Hopfield memories into smaller, continuous-time memories. Leveraging continuous attention, our new energy function modifies the update rule of HNs, replacing the traditional softmax-based probability mass function with a probability density, over the continuous memory. This formulation aligns with modern perspectives on human executive function, offering a principled link between attractor dynamics in working memory and resource-efficient memory allocation. Our framework maintains competitive performance with HNs while leveraging a compressed memory, reducing computational costs across synthetic and video datasets.
2502.10125
Learning Relational Tabular Data without Shared Features
cs.LG cs.AI
Learning relational tabular data has gained significant attention recently, but most studies focus on single tables, overlooking the potential of cross-table learning. Cross-table learning, especially in scenarios where tables lack shared features and pre-aligned data, offers vast opportunities but also introduces substantial challenges. The alignment space is immense, and determining accurate alignments between tables is highly complex. We propose Latent Entity Alignment Learning (Leal), a novel framework enabling effective cross-table training without requiring shared features or pre-aligned data. Leal operates on the principle that properly aligned data yield lower loss than misaligned data, a concept embodied in its soft alignment mechanism. This mechanism is coupled with a differentiable cluster sampler module, ensuring efficient scaling to large relational tables. Furthermore, we provide a theoretical proof of the cluster sampler's approximation capacity. Extensive experiments on five real-world and five synthetic datasets show that Leal achieves up to a 26.8% improvement in predictive performance compared to state-of-the-art methods, demonstrating its effectiveness and scalability.
2502.10127
Leveraging V2X for Collaborative HD Maps Construction Using Scene Graph Generation
cs.CV
High-Definition (HD) maps play a crucial role in autonomous vehicle navigation, complementing onboard perception sensors for improved accuracy and safety. Traditional HD map generation relies on dedicated mapping vehicles, which are costly and fail to capture real-time infrastructure changes. This paper presents HDMapLaneNet, a novel framework leveraging V2X communication and Scene Graph Generation to collaboratively construct a localized geometric layer of HD maps. The approach extracts lane centerlines from front-facing camera images, represents them as graphs, and transmits the data for global aggregation to the cloud via V2X. Preliminary results on the nuScenes dataset demonstrate superior association prediction performance compared to a state-of-the-art method.
2502.10138
Provably Efficient RL under Episode-Wise Safety in Constrained MDPs with Linear Function Approximation
cs.LG
We study the reinforcement learning (RL) problem in a constrained Markov decision process (CMDP), where an agent explores the environment to maximize the expected cumulative reward while satisfying a single constraint on the expected total utility value in every episode. While this problem is well understood in the tabular setting, theoretical results for function approximation remain scarce. This paper closes the gap by proposing an RL algorithm for linear CMDPs that achieves $\tilde{\mathcal{O}}(\sqrt{K})$ regret with an episode-wise zero-violation guarantee. Furthermore, our method is computationally efficient, scaling polynomially with problem-dependent parameters while remaining independent of the state space size. Our results significantly improve upon recent linear CMDP algorithms, which either violate the constraint or incur exponential computational costs.
2502.10140
Small Models, Big Impact: Efficient Corpus and Graph-Based Adaptation of Small Multilingual Language Models for Low-Resource Languages
cs.CL
Low-resource languages (LRLs) face significant challenges in natural language processing (NLP) due to limited data. While current state-of-the-art large language models (LLMs) still struggle with LRLs, smaller multilingual models (mLMs) such as mBERT and XLM-R offer greater promise due to a better fit of their capacity to low training data sizes. This study systematically investigates parameter-efficient adapter-based methods for adapting mLMs to LRLs, evaluating three architectures: Sequential Bottleneck, Invertible Bottleneck, and Low-Rank Adaptation. Using unstructured text from GlotCC and structured knowledge from ConceptNet, we show that small adaptation datasets (e.g., up to 1 GB of free-text or a few MB of knowledge graph data) yield gains in intrinsic (masked language modeling) and extrinsic tasks (topic classification, sentiment analysis, and named entity recognition). We find that Sequential Bottleneck adapters excel in language modeling, while Invertible Bottleneck adapters slightly outperform other methods on downstream tasks due to better embedding alignment and larger parameter counts. Adapter-based methods match or outperform full fine-tuning while using far fewer parameters, and smaller mLMs prove more effective for LRLs than massive LLMs like LLaMA-3, GPT-4, and DeepSeek-R1-based distilled models. While adaptation improves performance, pre-training data size remains the dominant factor, especially for languages with extensive pre-training coverage.
2502.10141
Pangraphs as models of higher-order interactions
physics.soc-ph cs.SI math.CO q-bio.MN q-bio.PE
Graphs depict pairwise relationships between objects within a system. Higher-order interactions (HOIs), which involve more than two objects simultaneously, are common in nature. Such interactions can change the stability of a complex system. Hypergraphs can represent an HOI as an arbitrary subset of vertices. However, they fail to capture the specific roles of the vertices involved, which can be highly asymmetric, particularly in the case of interaction modifications. We introduce pangraphs, a robust and quantitative generalisation of graphs that accurately captures arbitrarily complex higher-order interactions. We demonstrate that several higher-order representations proposed in the literature are specific instances of pangraphs. Additionally, we introduce an incidence multilayer digraph representation of a pangraph, referred to as Levi digraph. We adapt degree and Katz centrality measures to the pangraph framework and show that a consistent generalisation of recursive graph measures cannot be simplified to a Levi digraph of a pangraph. We construct a pangraph for a real-world coffee agroecosystem and compare Katz centrality between its dihypergraph and pangraph representations, both analytically and numerically. The choice of representation significantly affects centrality values and alters vertex ranks. Additionally, we emphasise the use of real-valued incidence matrices to quantify interaction strengths and the roles of vertices within the system.
2502.10145
Interpretable Concept-based Deep Learning Framework for Multimodal Human Behavior Modeling
cs.CV cs.MM
In the contemporary era of intelligent connectivity, Affective Computing (AC), which enables systems to recognize, interpret, and respond to human behavior states, has become an integrated part of many AI systems. As one of the most critical components of responsible AI and trustworthiness in all human-centered systems, explainability has been a major concern in AC. Particularly, the recently released EU General Data Protection Regulation requires any high-risk AI systems to be sufficiently interpretable, including biometric-based systems and emotion recognition systems widely used in the affective computing field. Existing explainable methods often compromise between interpretability and performance. Most of them focus only on highlighting key network parameters without offering meaningful, domain-specific explanations to the stakeholders. Additionally, they also face challenges in effectively co-learning and explaining insights from multimodal data sources. To address these limitations, we propose a novel and generalizable framework, namely the Attention-Guided Concept Model (AGCM), which provides learnable conceptual explanations by identifying what concepts that lead to the predictions and where they are observed. AGCM is extendable to any spatial and temporal signals through multimodal concept alignment and co-learning, empowering stakeholders with deeper insights into the model's decision-making process. We validate the efficiency of AGCM on well-established Facial Expression Recognition benchmark datasets while also demonstrating its generalizability on more complex real-world human behavior understanding applications.
2502.10148
Cooperative Multi-Agent Planning with Adaptive Skill Synthesis
cs.AI cs.MA
Despite much progress in training distributed artificial intelligence (AI), building cooperative multi-agent systems with multi-agent reinforcement learning (MARL) faces challenges in sample efficiency, interpretability, and transferability. Unlike traditional learning-based methods that require extensive interaction with the environment, large language models (LLMs) demonstrate remarkable capabilities in zero-shot planning and complex reasoning. However, existing LLM-based approaches heavily rely on text-based observations and struggle with the non-Markovian nature of multi-agent interactions under partial observability. We present COMPASS, a novel multi-agent architecture that integrates vision-language models (VLMs) with a dynamic skill library and structured communication for decentralized closed-loop decision-making. The skill library, bootstrapped from demonstrations, evolves via planner-guided tasks to enable adaptive strategies. COMPASS propagates entity information through multi-hop communication under partial observability. Evaluations on the improved StarCraft Multi-Agent Challenge (SMACv2) demonstrate COMPASS achieves up to 30\% higher win rates than state-of-the-art MARL algorithms in symmetric scenarios.
2502.10151
Semantica: Decentralized Search using a LLM-Guided Semantic Tree Overlay
cs.IR cs.DC cs.NI cs.SY eess.SY
Centralized search engines are key for the Internet, but lead to undesirable concentration of power. Decentralized alternatives fail to offer equal document retrieval accuracy and speed. Nevertheless, Semantic Overlay Networks can come close to the performance of centralized solutions when the semantics of documents are properly captured. This work uses embeddings from Large Language Models to capture semantics and fulfill the promise of Semantic Overlay Networks. Our proposed algorithm, called Semantica, constructs a prefix tree (trie) utilizing document embeddings calculated by a language model. Users connect to each other based on the embeddings of their documents, ensuring that semantically similar users are directly linked. Thereby, this construction makes it more likely for user searches to be answered by the users that they are directly connected to, or by the users they are close to in the network connection graph. The implementation of our algorithm also accommodates the semantic diversity of individual users by spawning "clone" user identifiers in the tree. Our experiments use emulation with a real-world workload to show Semantica's ability to identify and connect to similar users quickly. Semantica finds up to ten times more semantically similar users than current state-of-the-art approaches. At the same time, Semantica can retrieve more than two times the number of relevant documents given the same network load. We also make our code publicly available to facilitate further research in the area.
2502.10154
Video Soundtrack Generation by Aligning Emotions and Temporal Boundaries
cs.SD cs.AI cs.LG cs.MM eess.AS eess.IV
We introduce EMSYNC, a video-based symbolic music generation model that aligns music with a video's emotional content and temporal boundaries. It follows a two-stage framework, where a pretrained video emotion classifier extracts emotional features, and a conditional music generator produces MIDI sequences guided by both emotional and temporal cues. We introduce boundary offsets, a novel temporal conditioning mechanism that enables the model to anticipate and align musical chords with scene cuts. Unlike existing models, our approach retains event-based encoding, ensuring fine-grained timing control and expressive musical nuances. We also propose a mapping scheme to bridge the video emotion classifier, which produces discrete emotion categories, with the emotion-conditioned MIDI generator, which operates on continuous-valued valence-arousal inputs. In subjective listening tests, EMSYNC outperforms state-of-the-art models across all subjective metrics, for music theory-aware participants as well as the general listeners.
2502.10156
MonoForce: Learnable Image-conditioned Physics Engine
cs.RO cs.CV
We propose a novel model for the prediction of robot trajectories on rough offroad terrain from the onboard camera images. This model enforces the laws of classical mechanics through a physics-aware neural symbolic layer while preserving the ability to learn from large-scale data as it is end-to-end differentiable. The proposed hybrid model integrates a black-box component that predicts robot-terrain interaction forces with a neural-symbolic layer. This layer includes a differentiable physics engine that computes the robot's trajectory by querying these forces at the points of contact with the terrain. As the proposed architecture comprises substantial geometrical and physics priors, the resulting model can also be seen as a learnable physics engine conditioned on real images that delivers $10^4$ trajectories per second. We argue and empirically demonstrate that this architecture reduces the sim-to-real gap and mitigates out-of-distribution sensitivity. The differentiability, in conjunction with the rapid simulation speed, makes the model well-suited for various applications including model predictive control, trajectory shooting, supervised and reinforcement learning or SLAM. The codes and data are publicly available.
2502.10157
SessionRec: Next Session Prediction Paradigm For Generative Sequential Recommendation
cs.IR cs.AI
We introduce SessionRec, a novel next-session prediction paradigm (NSPP) for generative sequential recommendation, addressing the fundamental misalignment between conventional next-item prediction paradigm (NIPP) and real-world recommendation scenarios. Unlike NIPP's item-level autoregressive generation that contradicts actual session-based user interactions, our framework introduces a session-aware representation learning through hierarchical sequence aggregation (intra/inter-session), reducing attention computation complexity while enabling implicit modeling of massive negative interactions, and a session-based prediction objective that better captures users' diverse interests through multi-item recommendation in next sessions. Moreover, we found that incorporating a rank loss for items within the session under the next session prediction paradigm can significantly improve the ranking effectiveness of generative sequence recommendation models. We also verified that SessionRec exhibits clear power-law scaling laws similar to those observed in LLMs. Extensive experiments conducted on public datasets and online A/B test in Meituan App demonstrate the effectiveness of SessionRec. The proposed paradigm establishes new foundations for developing industrial-scale generative recommendation systems through its model-agnostic architecture and computational efficiency.
2502.10158
Combinatorial Reinforcement Learning with Preference Feedback
stat.ML cs.LG
In this paper, we consider combinatorial reinforcement learning with preference feedback, where a learning agent sequentially offers an action--an assortment of multiple items to--a user, whose preference feedback follows a multinomial logistic (MNL) model. This framework allows us to model real-world scenarios, particularly those involving long-term user engagement, such as in recommender systems and online advertising. However, this framework faces two main challenges: (1) the unknown value of each item, unlike traditional MNL bandits that only address single-step preference feedback, and (2) the difficulty of ensuring optimism while maintaining tractable assortment selection in the combinatorial action space with unknown values. In this paper, we assume a contextual MNL preference model, where the mean utilities are linear, and the value of each item is approximated by a general function. We propose an algorithm, MNL-VQL, that addresses these challenges, making it both computationally and statistically efficient. As a special case, for linear MDPs (with the MNL preference feedback), we establish the first regret lower bound in this framework and show that MNL-VQL achieves nearly minimax-optimal regret. To the best of our knowledge, this is the first work to provide statistical guarantees in combinatorial RL with preference feedback.
2502.10162
Revisiting Generalization Power of a DNN in Terms of Symbolic Interactions
cs.LG cs.AI cs.CL cs.CV
This paper aims to analyze the generalization power of deep neural networks (DNNs) from the perspective of interactions. Unlike previous analysis of a DNN's generalization power in a highdimensional feature space, we find that the generalization power of a DNN can be explained as the generalization power of the interactions. We found that the generalizable interactions follow a decay-shaped distribution, while non-generalizable interactions follow a spindle-shaped distribution. Furthermore, our theory can effectively disentangle these two types of interactions from a DNN. We have verified that our theory can well match real interactions in a DNN in experiments.
2502.10163
Enhancing anomaly detection with topology-aware autoencoders
hep-ph cs.LG hep-ex
Anomaly detection in high-energy physics is essential for identifying new physics beyond the Standard Model. Autoencoders provide a signal-agnostic approach but are limited by the topology of their latent space. This work explores topology-aware autoencoders, embedding phase-space distributions onto compact manifolds that reflect energy-momentum conservation. We construct autoencoders with spherical ($S^n$), product ($S^2 \otimes S^2$), and projective ($\mathbb{RP}^2$) latent spaces and compare their anomaly detection performance against conventional Euclidean embeddings. Our results show that autoencoders with topological priors significantly improve anomaly separation by preserving the global structure of the data manifold and reducing spurious reconstruction errors. Applying our approach to simulated hadronic top-quark decays, we show that latent spaces with appropriate topological constraints enhance sensitivity and robustness in detecting anomalous events. This study establishes topology-aware autoencoders as a powerful tool for unsupervised searches for new physics in particle-collision data.
2502.10173
Agentic End-to-End De Novo Protein Design for Tailored Dynamics Using a Language Diffusion Model
q-bio.BM cond-mat.mes-hall cond-mat.mtrl-sci cs.LG
Proteins are dynamic molecular machines whose biological functions, spanning enzymatic catalysis, signal transduction, and structural adaptation, are intrinsically linked to their motions. Designing proteins with targeted dynamic properties, however, remains a challenge due to the complex, degenerate relationships between sequence, structure, and molecular motion. Here, we introduce VibeGen, a generative AI framework that enables end-to-end de novo protein design conditioned on normal mode vibrations. VibeGen employs an agentic dual-model architecture, comprising a protein designer that generates sequence candidates based on specified vibrational modes and a protein predictor that evaluates their dynamic accuracy. This approach synergizes diversity, accuracy, and novelty during the design process. Via full-atom molecular simulations as direct validation, we demonstrate that the designed proteins accurately reproduce the prescribed normal mode amplitudes across the backbone while adopting various stable, functionally relevant structures. Notably, generated sequences are de novo, exhibiting no significant similarity to natural proteins, thereby expanding the accessible protein space beyond evolutionary constraints. Our work integrates protein dynamics into generative protein design, and establishes a direct, bidirectional link between sequence and vibrational behavior, unlocking new pathways for engineering biomolecules with tailored dynamical and functional properties. This framework holds broad implications for the rational design of flexible enzymes, dynamic scaffolds, and biomaterials, paving the way toward dynamics-informed AI-driven protein engineering.
2502.10174
Technical Risks of (Lethal) Autonomous Weapons Systems
cs.CY cs.AI cs.SY eess.SY
The autonomy and adaptability of (Lethal) Autonomous Weapons Systems, (L)AWS in short, promise unprecedented operational capabilities, but they also introduce profound risks that challenge the principles of control, accountability, and stability in international security. This report outlines the key technological risks associated with (L)AWS deployment, emphasizing their unpredictability, lack of transparency, and operational unreliability, which can lead to severe unintended consequences. Key Takeaways: 1. Proposed advantages of (L)AWS can only be achieved through objectification and classification, but a range of systematic risks limit the reliability and predictability of classifying algorithms. 2. These systematic risks include the black-box nature of AI decision-making, susceptibility to reward hacking, goal misgeneralization and potential for emergent behaviors that escape human control. 3. (L)AWS could act in ways that are not just unexpected but also uncontrollable, undermining mission objectives and potentially escalating conflicts. 4. Even rigorously tested systems may behave unpredictably and harmfully in real-world conditions, jeopardizing both strategic stability and humanitarian principles.
2502.10177
STMA: A Spatio-Temporal Memory Agent for Long-Horizon Embodied Task Planning
cs.AI
A key objective of embodied intelligence is enabling agents to perform long-horizon tasks in dynamic environments while maintaining robust decision-making and adaptability. To achieve this goal, we propose the Spatio-Temporal Memory Agent (STMA), a novel framework designed to enhance task planning and execution by integrating spatio-temporal memory. STMA is built upon three critical components: (1) a spatio-temporal memory module that captures historical and environmental changes in real time, (2) a dynamic knowledge graph that facilitates adaptive spatial reasoning, and (3) a planner-critic mechanism that iteratively refines task strategies. We evaluate STMA in the TextWorld environment on 32 tasks, involving multi-step planning and exploration under varying levels of complexity. Experimental results demonstrate that STMA achieves a 31.25% improvement in success rate and a 24.7% increase in average score compared to the state-of-the-art model. The results highlight the effectiveness of spatio-temporal memory in advancing the memory capabilities of embodied agents.
2502.10178
From Markov to Laplace: How Mamba In-Context Learns Markov Chains
cs.LG cs.AI cs.IT math.IT
While transformer-based language models have driven the AI revolution thus far, their computational complexity has spurred growing interest in viable alternatives, such as structured state space sequence models (SSMs) and Selective SSMs. Among these, Mamba (S6) and its variant Mamba-2 have shown remarkable inference speed ups over transformers while achieving comparable or superior performance on complex language modeling tasks. However, despite these architectural innovations and empirical successes, the fundamental learning capabilities of Mamba remain poorly understood. In this paper, we address this gap by studying in-context learning (ICL) on Markov chains and uncovering a surprising phenomenon: unlike transformers, even a single-layer Mamba efficiently learns the in-context Laplacian smoothing estimator, which is both Bayes and minimax optimal, for all Markovian orders. To explain this, we theoretically characterize the representation capacity of Mamba and reveal the fundamental role of convolution in enabling it to represent the optimal Laplacian smoothing. These theoretical insights align strongly with empirical results and, to the best of our knowledge, represent the first formal connection between Mamba and optimal statistical estimators. Finally, we outline promising research directions inspired by these findings.
2502.10180
Safe platooning control of connected and autonomous vehicles on curved multi-lane roads
eess.SY cs.SY
This paper investigates the safe platoon formation tracking and merging control problem of connected and automated vehicles (CAVs) on curved multi-lane roads. The first novelty is the separation of the control designs into two distinct parts: a lateral control law that ensures a geometrical convergence towards the reference path regardless of the translational velocity, and a longitudinal control design for each vehicle to achieve the desired relative arc length and velocity with respect to its neighboring vehicle. The second novelty is exploiting the constructive barrier feedback as an additive term to the nominal tracking control, ensuring both lateral and longitudinal collision avoidance. This constructive barrier feedback acts as a dissipative term, slowing down the relative velocity toward obstacles without affecting the nominal controller's performance. Consequently, our proposed control method enables safe platoon formation of vehicles on curved multi-lane roads, with theoretical guarantees for safety invariance and stability analysis. Simulation and experimental results on connected vehicles are provided to further validate the effectiveness of the proposed method.
2502.10183
Doing More With Less: Towards More Data-Efficient Syndrome-Based Neural Decoders
cs.IT math.IT
While significant research efforts have been directed toward developing more capable neural decoding architectures, comparatively little attention has been paid to the quality of training data. In this study, we address the challenge of constructing effective training datasets to maximize the potential of existing syndrome-based neural decoder architectures. We emphasize the advantages of using fixed datasets over generating training data dynamically and explore the problem of selecting appropriate training targets within this framework. Furthermore,we propose several heuristics for selecting training samples and present experimental evidence demonstrating that, with carefully curated datasets, it is possible to train neural decoders to achieve superior performance while requiring fewer training examples.
2502.10184
Realistic Evaluation of Deep Partial-Label Learning Algorithms
cs.LG
Partial-label learning (PLL) is a weakly supervised learning problem in which each example is associated with multiple candidate labels and only one is the true label. In recent years, many deep PLL algorithms have been developed to improve model performance. However, we find that some early developed algorithms are often underestimated and can outperform many later algorithms with complicated designs. In this paper, we delve into the empirical perspective of PLL and identify several critical but previously overlooked issues. First, model selection for PLL is non-trivial, but has never been systematically studied. Second, the experimental settings are highly inconsistent, making it difficult to evaluate the effectiveness of the algorithms. Third, there is a lack of real-world image datasets that can be compatible with modern network architectures. Based on these findings, we propose PLENCH, the first Partial-Label learning bENCHmark to systematically compare state-of-the-art deep PLL algorithms. We investigate the model selection problem for PLL for the first time, and propose novel model selection criteria with theoretical guarantees. We also create Partial-Label CIFAR-10 (PLCIFAR10), an image dataset of human-annotated partial labels collected from Amazon Mechanical Turk, to provide a testbed for evaluating the performance of PLL algorithms in more realistic scenarios. Researchers can quickly and conveniently perform a comprehensive and fair evaluation and verify the effectiveness of newly developed algorithms based on PLENCH. We hope that PLENCH will facilitate standardized, fair, and practical evaluation of PLL algorithms in the future.
2502.10185
A Powerful Random Forest Featuring Linear Extensions (RaFFLE)
cs.LG
Random forests are widely used in regression. However, the decision trees used as base learners are poor approximators of linear relationships. To address this limitation we propose RaFFLE (Random Forest Featuring Linear Extensions), a novel framework that integrates the recently developed PILOT trees (Piecewise Linear Organic Trees) as base learners within a random forest ensemble. PILOT trees combine the computational efficiency of traditional decision trees with the flexibility of linear model trees. To ensure sufficient diversity of the individual trees, we introduce an adjustable regularization parameter and use node-level feature sampling. These modifications improve the accuracy of the forest. We establish theoretical guarantees for the consistency of RaFFLE under weak conditions, and its faster convergence when the data are generated by a linear model. Empirical evaluations on 136 regression datasets demonstrate that RaFFLE outperforms the classical CART and random forest methods, the regularized linear methods Lasso and Ridge, and the state-of-the-art XGBoost algorithm, across both linear and nonlinear datasets. By balancing predictive accuracy and computational efficiency, RaFFLE proves to be a versatile tool for tackling a wide variety of regression problems.
2502.10187
Reinforcement Learning based Constrained Optimal Control: an Interpretable Reward Design
eess.SY cs.SY
This paper presents an interpretable reward design framework for reinforcement learning based constrained optimal control problems with state and terminal constraints. The problem is formalized within a standard partially observable Markov decision process framework. The reward function is constructed from four weighted components: a terminal constraint reward, a guidance reward, a penalty for state constraint violations, and a cost reduction incentive reward. A theoretically justified reward design is then presented, which establishes bounds on the weights of the components. This approach ensures that constraints are satisfied and objectives are optimized while mitigating numerical instability. Acknowledging the importance of prior knowledge in reward design, we sequentially solve two subproblems, using each solution to inform the reward design for the subsequent problem. Subsequently, we integrate reinforcement learning with curriculum learning, utilizing policies derived from simpler subproblems to assist in tackling more complex challenges, thereby facilitating convergence. The framework is evaluated against original and randomly weighted reward designs in a multi-agent particle environment. Experimental results demonstrate that the proposed approach significantly enhances satisfaction of terminal and state constraints and optimization of control cost.
2502.10192
A Note on "Constructing Bent Functions Outside the Maiorana-McFarland Class Using a General Form of Rothaus"
cs.IT math.IT
In 2017, Zhang et al. proposed a question (not open problem) and two open problems in [IEEE TIT 63 (8): 5336--5349, 2017] about constructing bent functions by using Rothaus' construction. In this note, we prove that the sufficient conditions of Rothaus' construction are also necessary, which answers their question. Besides, we demonstrate that the second open problem, which considers the iterative method of constructing bent functions by using Rothaus' construction, has only a trivial solution. It indicates that all bent functions obtained by using Rothaus' construction iteratively can be generated from the direct sum of an initial bent function and a quadratic bent function. This directly means that Zhang et al.'s construction idea makes no contribution to the construction of bent functions. To compensate the weakness of their work, we propose an iterative construction of bent functions by using a secondary construction in [DCC 88: 2007--2035, 2020].
2502.10193
Merging public elementary schools to reduce racial/ethnic segregation
cs.CY cs.AI
Diverse schools can help address implicit biases and increase empathy, mutual respect, and reflective thought by fostering connections between students from different racial/ethnic, socioeconomic, and other backgrounds. Unfortunately, demographic segregation remains rampant in US public schools, despite over 70 years since the passing of federal legislation formally outlawing segregation by race. However, changing how students are assigned to schools can help foster more integrated learning environments. In this paper, we explore "school mergers" as one such under-explored, yet promising, student assignment policy change. School mergers involve merging the school attendance boundaries, or catchment areas, of schools and subsequently changing the grades each school offers. We develop an algorithm to simulate elementary school mergers across 200 large school districts serving 4.5 million elementary school students and find that pairing or tripling schools in this way could reduce racial/ethnic segregation by a median relative 20% -- and as much as nearly 60% in some districts -- while increasing driving times to schools by an average of a few minutes each way. Districts with many interfaces between racially/ethnically-disparate neighborhoods tend to be prime candidates for mergers. We also compare the expected results of school mergers to other typical integration policies, like redistricting, and find that different policies may be more or less suitable in different places. Finally, we make our results available through a public dashboard for policymakers and community members to explore further (https://mergers.schooldiversity.org). Together, our study offers new findings and tools to support integration policy-making across US public school districts.
2502.10195
Exploring the Camera Bias of Person Re-identification
cs.CV cs.AI cs.LG
We empirically investigate the camera bias of person re-identification (ReID) models. Previously, camera-aware methods have been proposed to address this issue, but they are largely confined to training domains of the models. We measure the camera bias of ReID models on unseen domains and reveal that camera bias becomes more pronounced under data distribution shifts. As a debiasing method for unseen domain data, we revisit feature normalization on embedding vectors. While the normalization has been used as a straightforward solution, its underlying causes and broader applicability remain unexplored. We analyze why this simple method is effective at reducing bias and show that it can be applied to detailed bias factors such as low-level image properties and body angle. Furthermore, we validate its generalizability across various models and benchmarks, highlighting its potential as a simple yet effective test-time postprocessing method for ReID. In addition, we explore the inherent risk of camera bias in unsupervised learning of ReID models. The unsupervised models remain highly biased towards camera labels even for seen domain data, indicating substantial room for improvement. Based on observations of the negative impact of camera-biased pseudo labels on training, we suggest simple training strategies to mitigate the bias. By applying these strategies to existing unsupervised learning algorithms, we show that significant performance improvements can be achieved with minor modifications.
2502.10197
MathConstruct: Challenging LLM Reasoning with Constructive Proofs
cs.AI
While Large Language Models (LLMs) demonstrate impressive performance in mathematics, existing math benchmarks come with significant limitations. Many focus on problems with fixed ground-truth answers, and are often saturated due to problem simplicity or the viability of guessing or memorization. Crucially, they capture only a narrow subset of relevant math problems. To address this research gap, we introduce \mc, a new benchmark of 126 challenging problems sourced from various math competitions, which targets constructive proofs, a widely encountered problem type requiring the construction of mathematical objects with specific properties. These proofs are particularly suitable for LLM evaluation, as solution correctness can be easily verified. Our automated verifiers also enable MathConstruct to generate problem variations, used to evaluate robustness. State-of-the-art LLMs solve only 54% of MathConstruct problems, highlighting its complexity and importance for LLM evaluation.
2502.10200
Dynamic Reinforcement Learning for Actors
cs.LG cs.AI cs.NE
Dynamic Reinforcement Learning (Dynamic RL), proposed in this paper, directly controls system dynamics, instead of the actor (action-generating neural network) outputs at each moment, bringing about a major qualitative shift in reinforcement learning (RL) from static to dynamic. The actor is initially designed to generate chaotic dynamics through the loop with its environment, enabling the agent to perform flexible and deterministic exploration. Dynamic RL controls global system dynamics using a local index called "sensitivity," which indicates how much the input neighborhood contracts or expands into the corresponding output neighborhood through each neuron's processing. While sensitivity adjustment learning (SAL) prevents excessive convergence of the dynamics, sensitivity-controlled reinforcement learning (SRL) adjusts them -- to converge more to improve reproducibility around better state transitions with positive TD error and to diverge more to enhance exploration around worse transitions with negative TD error. Dynamic RL was applied only to the actor in an Actor-Critic RL architecture while applying it to the critic remains a challenge. It was tested on two dynamic tasks and functioned effectively without external exploration noise or backward computation through time. Moreover, it exhibited excellent adaptability to new environments, although some problems remain. Drawing parallels between 'exploration' and 'thinking,' the author hypothesizes that "exploration grows into thinking through learning" and believes this RL could be a key technique for the emergence of thinking, including inspiration that cannot be reconstructed from massive existing text data. Finally, despite being presumptuous, the author presents the argument that this research should not proceed due to its potentially fatal risks, aiming to encourage discussion.
2502.10201
Prediction hubs are context-informed frequent tokens in LLMs
cs.CL cs.AI
Hubness, the tendency for few points to be among the nearest neighbours of a disproportionate number of other points, commonly arises when applying standard distance measures to high-dimensional data, often negatively impacting distance-based analysis. As autoregressive large language models (LLMs) operate on high-dimensional representations, we ask whether they are also affected by hubness. We first show, theoretically, that the only representation comparison operation performed by LLMs, namely that between context and unembedding vectors to determine continuation probabilities, is not characterized by the concentration of distances phenomenon that typically causes the appeareance of nuisance hubness. We then empirically show that this comparison still leads to a high degree of hubness, but the hubs in this case do not constitute a disturbance. They are rather the result of context-modulated frequent tokens often appearing in the pool of likely candidates for next token prediction. On the other hand, when other distance computations involving LLM representations are performed, we do not have the same theoretical guarantees, and, indeed, we see nuisance hubs appear. In summary, our work highlights, on the one hand, how hubness, while omnipresent in high-dimensional spaces, is not always a negative property that needs to be mitigated, and, on the other hand, it shows that various widely-used LLMs have developed a guessing strategy that consists in constantly assigning a high probability to frequent tokens.
2502.10202
Can Post-Training Quantization Benefit from an Additional QLoRA Integration?
cs.CL
Large language models (LLMs) have transformed natural language processing but pose significant challenges for real-world deployment. These models necessitate considerable computing resources, which can be costly and frequently unavailable. Model compression techniques such as quantization are often leveraged to alleviate resource demand, but they may have a negative impact on the generation quality. In this study, we explore the integration of 4-bit Post-training Quantization (PTQ) with QLoRA to address these issues. We demonstrate through extensive experiments that this integration outperforms standard PTQ, and in some cases even 16-bit full-parameter fine-tuning on LLMs, validated across proprietary and public datasets with different quantization algorithms. The results demonstrate the efficacy of PTQ-QLoRA integration, offering a viable solution for deploying powerful LLMs in resource-constrained environments without compromising on performance.
2502.10203
AI-in-the-Loop Sensing and Communication Joint Design for Edge Intelligence
cs.LG cs.DC
Recent breakthroughs in artificial intelligence (AI), wireless communications, and sensing technologies have accelerated the evolution of edge intelligence. However, conventional systems still grapple with issues such as low communication efficiency, redundant data acquisition, and poor model generalization. To overcome these challenges, we propose an innovative framework that enhances edge intelligence through AI-in-the-loop joint sensing and communication (JSAC). This framework features an AI-driven closed-loop control architecture that jointly optimizes system resources, thereby delivering superior system-level performance. A key contribution of our work is establishing an explicit relationship between validation loss and the system's tunable parameters. This insight enables dynamic reduction of the generalization error through AI-driven closed-loop control. Specifically, for sensing control, we introduce an adaptive data collection strategy based on gradient importance sampling, allowing edge devices to autonomously decide when to terminate data acquisition and how to allocate sample weights based on real-time model feedback. For communication control, drawing inspiration from stochastic gradient Langevin dynamics (SGLD), our joint optimization of transmission power and batch size converts channel and data noise into gradient perturbations that help mitigate overfitting. Experimental evaluations demonstrate that our framework reduces communication energy consumption by up to 77 percent and sensing costs measured by the number of collected samples by up to 52 percent while significantly improving model generalization -- with up to 58 percent reductions of the final validation loss. It validates that the proposed scheme can harvest the mutual benefit of AI and JSAC systems by incorporating the model itself into the control loop of the system.
2502.10205
Looking around you: external information enhances representations for event sequences
cs.LG
Representation learning produces models in different domains, such as store purchases, client transactions, and general people's behaviour. However, such models for sequential data usually process a single sequence, ignoring context from other relevant ones, even in domains with rapidly changing external environments like finance or misguiding the prediction for a user with no recent events. We are the first to propose a method that aggregates information from multiple user representations augmenting a specific user one for a scenario of multiple co-occurring event sequences. Our study considers diverse aggregation approaches, ranging from simple pooling techniques to trainable attention-based approaches, especially Kernel attention aggregation, that can highlight more complex information flow from other users. The proposed method operates atop an existing encoder and supports its efficient fine-tuning. Across considered datasets of financial transactions and downstream tasks, Kernel attention improves ROC AUC scores, both with and without fine-tuning, while mean pooling yields a smaller but still significant gain.
2502.10207
RIPOST: Two-Phase Private Decomposition for Multidimensional Data
cs.DB
Differential privacy (DP) is considered as the gold standard for data privacy. While the problem of answering simple queries and functions under DP guarantees has been thoroughly addressed in recent years, the problem of releasing multidimensional data under DP remains challenging. In this paper, we focus on this problem, in particular on how to construct privacy-preserving views using a domain decomposition approach. The main idea is to recursively split the domain into sub-domains until a convergence condition is met. The resulting sub-domains are perturbed and then published in order to be used to answer arbitrary queries. Existing methods that have addressed this problem using domain decomposition face two main challenges: (i) efficient privacy budget management over a variable and undefined decomposition depth $h$; and (ii) defining an optimal data-dependent splitting strategy that minimizes the error in the sub-domains while ensuring the smallest possible decomposition. To address these challenges, we present RIPOST, a multidimensional data decomposition algorithm that bypasses the constraint of predefined depth $h$ and applies a data-aware splitting strategy to optimize the quality of the decomposition results.The core of RIPOST is a two-phase strategy that separates non-empty sub-domains at an early stage from empty sub-domains by exploiting the properties of multidimensional datasets, and then decomposes the resulting sub-domains with minimal inaccuracies using the mean function. Moreover, RIPOST introduces a privacy budget distribution that allows decomposition without requiring prior computation of the depth $h$. Through extensive experiments, we demonstrated that \texttt{RIPOST} outperforms state-of-the-art methods in terms of data utility and accuracy on a variety of datasets and test cases
2502.10208
SGS-GNN: A Supervised Graph Sparsification method for Graph Neural Networks
cs.LG
We propose SGS-GNN, a novel supervised graph sparsifier that learns the sampling probability distribution of edges and samples sparse subgraphs of a user-specified size to reduce the computational costs required by GNNs for inference tasks on large graphs. SGS-GNN employs regularizers in the loss function to enhance homophily in sparse subgraphs, boosting the accuracy of GNNs on heterophilic graphs, where a significant number of the neighbors of a node have dissimilar labels. SGS-GNN also supports conditional updates of the probability distribution learning module based on a prior, which helps narrow the search space for sparse graphs. SGS-GNN requires fewer epochs to obtain high accuracies since it learns the search space of subgraphs more effectively than methods using fixed distributions such as random sampling. Extensive experiments using 33 homophilic and heterophilic graphs demonstrate the following: (i) with only 20% of edges retained in the sparse subgraphs, SGS-GNN improves the F1-scores by a geometric mean of 4% relative to the original graph; on heterophilic graphs, the prediction accuracy is better up to 30%. (ii) SGS-GNN outperforms state-of-the-art methods with improvement in F1-scores of 4-7% in geometric mean with similar sparsities in the sampled subgraphs, and (iii) compared to sparsifiers that employ fixed distributions, SGS-GNN requires about half the number of epochs to converge.
2502.10209
Mutual Coupling in Holographic MIMO: Physical Modeling and Information-Theoretic Analysis
cs.IT math.IT
This paper presents a comprehensive framework for holographic multiantenna communication, a paradigm that integrates both wide apertures and closely spaced antennas relative to the wavelength. The presented framework is physically grounded, enabling information-theoretic analyses that inherently incorporate correlation and mutual coupling among the antennas. This establishes the combined effects of correlation and coupling on the information-theoretic performance limits across SNR levels. Additionally, it reveals that, by suitably selecting the individual antenna patterns, mutual coupling can be harnessed to either reinforce or counter spatial correlations as appropriate for specific SNRs, thereby improving the performance.
2502.10211
Control-flow anomaly detection by process mining-based feature extraction and dimensionality reduction
cs.LG
The business processes of organizations may deviate from normal control flow due to disruptive anomalies, including unknown, skipped, and wrongly-ordered activities. To identify these control-flow anomalies, process mining can check control-flow correctness against a reference process model through conformance checking, an explainable set of algorithms that allows linking any deviations with model elements. However, the effectiveness of conformance checking-based techniques is negatively affected by noisy event data and low-quality process models. To address these shortcomings and support the development of competitive and explainable conformance checking-based techniques for control-flow anomaly detection, we propose a novel process mining-based feature extraction approach with alignment-based conformance checking. This variant aligns the deviating control flow with a reference process model; the resulting alignment can be inspected to extract additional statistics such as the number of times a given activity caused mismatches. We integrate this approach into a flexible and explainable framework for developing techniques for control-flow anomaly detection. The framework combines process mining-based feature extraction and dimensionality reduction to handle high-dimensional feature sets, achieve detection effectiveness, and support explainability. The results show that the framework techniques implementing our approach outperform the baseline conformance checking-based techniques while maintaining the explainable nature of conformance checking. We also provide an explanation of why existing conformance checking-based techniques may be ineffective.
2502.10214
Mapping bathymetry of inland water bodies on the North Slope of Alaska with Landsat using Random Forest
cs.CV cs.LG
The North Slope of Alaska is dominated by small waterbodies that provide critical ecosystem services for local population and wildlife. Detailed information on the depth of the waterbodies is scarce due to the challenges with collecting such information. In this work we have trained a machine learning (Random Forest Regressor) model to predict depth from multispectral Landsat data in waterbodies across the North Slope of Alaska. The greatest challenge is the scarcity of in situ data, which is expensive and difficult to obtain, to train the model. We overcame this challenge by using modeled depth predictions from a prior study as synthetic training data to provide a more diverse training data pool for the Random Forest. The final Random Forest model was more robust than models trained directly on the in situ data and when applied to 208 Landsat 8 scenes from 2016 to 2018 yielded a map with an overall $r^{2}$ value of 0.76 on validation. The final map has been made available through the Oak Ridge National Laboratory Distribute Active Archive Center (ORNL-DAAC). This map represents a first of its kind regional assessment of waterbody depth with per pixel estimates of depth for the entire North Slope of Alaska.
2502.10215
Do Large Language Models Reason Causally Like Us? Even Better?
cs.AI cs.LG
Causal reasoning is a core component of intelligence. Large language models (LLMs) have shown impressive capabilities in generating human-like text, raising questions about whether their responses reflect true understanding or statistical patterns. We compared causal reasoning in humans and four LLMs using tasks based on collider graphs, rating the likelihood of a query variable occurring given evidence from other variables. We find that LLMs reason causally along a spectrum from human-like to normative inference, with alignment shifting based on model, context, and task. Overall, GPT-4o and Claude showed the most normative behavior, including "explaining away", whereas Gemini-Pro and GPT-3.5 did not. Although all agents deviated from the expected independence of causes - Claude the least - they exhibited strong associative reasoning and predictive inference when assessing the likelihood of the effect given its causes. These findings underscore the need to assess AI biases as they increasingly assist human decision-making.
2502.10216
Forget the Data and Fine-Tuning! Just Fold the Network to Compress
cs.LG cs.AI
We introduce model folding, a novel data-free model compression technique that merges structurally similar neurons across layers, significantly reducing the model size without the need for fine-tuning or access to training data. Unlike existing methods, model folding preserves data statistics during compression by leveraging k-means clustering, and using novel data-free techniques to prevent variance collapse or explosion. Our theoretical framework and experiments across standard benchmarks, including ResNet18 and LLaMA-7B, demonstrate that model folding achieves comparable performance to data-driven compression techniques and outperforms recently proposed data-free methods, especially at high sparsity levels. This approach is particularly effective for compressing large-scale models, making it suitable for deployment in resource-constrained environments.
2502.10218
Integrated Multi-Simulation Environments for Aerial Robotics Research
cs.RO
Simulation frameworks play a pivotal role in the safe development of robotic applications. However, often different components of an envisioned robotic system are best simulated in different environments/simulators. This poses a significant challenge in simulating the entire project into a single integrated robotic framework. Specifically, for partially-open or closed-source simulators, often two core limitations arise. i) Actors in the scene other than the designated robots cannot be controlled during runtime via interfaces such as ROS and ii) retrieving real-time state information (such as pose, velocity etc.) of objects in the scene is prevented. In this work, we address these limitations and describe our solution for the use case of integrating aerial drones simulated by the powerful simulator Sphinx (provided by Parrot Drone) into the Gazebo simulator. We achieve this by means of a mirrored instance of a drone that is included into existing Gazebo-based environments. A promising application of our integrated simulation environment is the task of target tracking that is common in aerial multi-robot scenarios. Therefore, to demonstrate the effectiveness our our integrated simulation, we also implement a model predictive controller (MPC) that outperforms the default PID-based controller framework provided with the Parrot's popular Anafi drone in various dynamic tracking scenarios thus enhancing the utility of the overall system. We test our solution by including the Anafi drone in an existing Gazebo-based simulation and evaluate the performance of the MPC through rigorous testing in simulated and real-world tracking experiments against a customized PID controller baseline. Source code is published on https://github.com/robot-perception-group/anafi_sim.
2502.10220
Optimal and Coordinated Voltage Control: Case Study on a 132 kV Norwegian Grid Subsystem
eess.SY cs.SY
This work presents a framework for dynamic performance assessment of the higher layers in the hierarchical voltage regulation scheme, with case studies applied to specific areas of the Norwegian grid. Unlike the primary (PVR) level, the secondary (SVR) and tertiary (TVR) levels are not tuned to a single device at a time, handling instead several reactive power resources available within a control zone including generator units, static VAr compensators and others. Proper SVR-TVR coordination for realistic transmission systems is a challenging topic at the core of many ongoing discussions in voltage control literature. Special focus is placed on practical considerations from the system operator perspective, since this research is also aimed at simplifying daily control centre routines. Dynamic simulation results concern a 21-bus equivalent of a 132 kV network model that accurately represents a Norwegian grid subsystem. Case studies address daily grid operation with real-life load demand and wind power generation profiles, showing that the proposed strategy is effective not only to minimize total active power losses as much as possible within system-wide limitations, but also to maintain adequate voltage profiles and reactive power flows. Findings pertaining to this work showcase the benefits of applying hierarchical voltage regulation layers as an asset to day-to-day control center management of a realistic transmission network.
2502.10224
Comparison of Deep Recurrent Neural Networks and Bayesian Neural Networks for Detecting Electric Motor Damage Through Sound Signal Analysis
cs.LG
Fault detection in electric motors is a critical challenge in various industries, where failures can result in significant operational disruptions. This study investigates the use of Recurrent Neural Networks (RNNs) and Bayesian Neural Networks (BNNs) for diagnosing motor damage using acoustic signal analysis. A novel approach is proposed, leveraging frequency domain representation of sound signals for enhanced diagnostic accuracy. The architectures of both RNNs and BNNs are designed and evaluated on real-world acoustic data collected from household appliances using smartphones. Experimental results demonstrate that BNNs provide superior fault detection performance, particularly for imbalanced datasets, offering more robust and interpretable predictions compared to traditional methods. The findings suggest that BNNs, with their ability to incorporate uncertainty, are well-suited for industrial diagnostic applications. Further analysis and benchmarks are suggested to explore resource efficiency and classification capabilities of these architectures.
2502.10226
A Multiagent Path Search Algorithm for Large-Scale Coalition Structure Generation
cs.MA cs.AI cs.GT
Coalition structure generation (CSG), i.e. the problem of optimally partitioning a set of agents into coalitions to maximize social welfare, is a fundamental computational problem in multiagent systems. This problem is important for many applications where small run times are necessary, including transportation and disaster response. In this paper, we develop SALDAE, a multiagent path finding algorithm for CSG that operates on a graph of coalition structures. Our algorithm utilizes a variety of heuristics and strategies to perform the search and guide it. It is an anytime algorithm that can handle large problems with hundreds and thousands of agents. We show empirically on nine standard value distributions, including disaster response and electric vehicle allocation benchmarks, that our algorithm enables a rapid finding of high-quality solutions and compares favorably with other state-of-the-art methods.
2502.10230
ProReco: A Process Discovery Recommender System
cs.LG cs.IR
Process discovery aims to automatically derive process models from historical execution data (event logs). While various process discovery algorithms have been proposed in the last 25 years, there is no consensus on a dominating discovery algorithm. Selecting the most suitable discovery algorithm remains a challenge due to competing quality measures and diverse user requirements. Manually selecting the most suitable process discovery algorithm from a range of options for a given event log is a time-consuming and error-prone task. This paper introduces ProReco, a Process discovery Recommender system designed to recommend the most appropriate algorithm based on user preferences and event log characteristics. ProReco incorporates state-of-the-art discovery algorithms, extends the feature pools from previous work, and utilizes eXplainable AI (XAI) techniques to provide explanations for its recommendations.
2502.10233
Learning to Solve the Min-Max Mixed-Shelves Picker-Routing Problem via Hierarchical and Parallel Decoding
cs.MA cs.LG stat.ML
The Mixed-Shelves Picker Routing Problem (MSPRP) is a fundamental challenge in warehouse logistics, where pickers must navigate a mixed-shelves environment to retrieve SKUs efficiently. Traditional heuristics and optimization-based approaches struggle with scalability, while recent machine learning methods often rely on sequential decision-making, leading to high solution latency and suboptimal agent coordination. In this work, we propose a novel hierarchical and parallel decoding approach for solving the min-max variant of the MSPRP via multi-agent reinforcement learning. While our approach generates a joint distribution over agent actions, allowing for fast decoding and effective picker coordination, our method introduces a sequential action selection to avoid conflicts in the multi-dimensional action space. Experiments show state-of-the-art performance in both solution quality and inference speed, particularly for large-scale and out-of-distribution instances. Our code is publicly available at http://github.com/LTluttmann/marl4msprp.
2502.10235
AdaPTS: Adapting Univariate Foundation Models to Probabilistic Multivariate Time Series Forecasting
stat.ML cs.LG
Pre-trained foundation models (FMs) have shown exceptional performance in univariate time series forecasting tasks. However, several practical challenges persist, including managing intricate dependencies among features and quantifying uncertainty in predictions. This study aims to tackle these critical limitations by introducing adapters; feature-space transformations that facilitate the effective use of pre-trained univariate time series FMs for multivariate tasks. Adapters operate by projecting multivariate inputs into a suitable latent space and applying the FM independently to each dimension. Inspired by the literature on representation learning and partially stochastic Bayesian neural networks, we present a range of adapters and optimization/inference strategies. Experiments conducted on both synthetic and real-world datasets confirm the efficacy of adapters, demonstrating substantial enhancements in forecasting accuracy and uncertainty quantification compared to baseline methods. Our framework, AdaPTS, positions adapters as a modular, scalable, and effective solution for leveraging time series FMs in multivariate contexts, thereby promoting their wider adoption in real-world applications. We release the code at https://github.com/abenechehab/AdaPTS.
2502.10236
Shaping Inductive Bias in Diffusion Models through Frequency-Based Noise Control
cs.LG cs.AI
Diffusion Probabilistic Models (DPMs) are powerful generative models that have achieved unparalleled success in a number of generative tasks. In this work, we aim to build inductive biases into the training and sampling of diffusion models to better accommodate the target distribution of the data to model. For topologically structured data, we devise a frequency-based noising operator to purposefully manipulate, and set, these inductive biases. We first show that appropriate manipulations of the noising forward process can lead DPMs to focus on particular aspects of the distribution to learn. We show that different datasets necessitate different inductive biases, and that appropriate frequency-based noise control induces increased generative performance compared to standard diffusion. Finally, we demonstrate the possibility of ignoring information at particular frequencies while learning. We show this in an image corruption and recovery task, where we train a DPM to recover the original target distribution after severe noise corruption.
2502.10239
Efficient Zero-Order Federated Finetuning of Language Models for Resource-Constrained Devices
cs.LG cs.AI
Federated fine-tuning offers a promising approach for tuning Large Language Models (LLMs) on edge devices while preserving data privacy. However, fine-tuning these models on edge devices remains challenging due to high memory, communication, and computational demands. Zero-order optimization with task alignment provides a potential solution, enabling fine-tuning with inference-level memory requirements but requires a longer convergence time. In this paper, we propose Federated Split-Perturbation Zero-order Optimization (FedSPZO) that divides the network into two blocks, applying a different number of perturbations per block in a computationally effective way, achieving faster convergence. Our evaluation shows a $2.5 - 7\times $ reduction in computation overhead compared to zero-order state of the art techniques in federated learning.
2502.10243
Safety Blind Spot in Remote Driving: Considerations for Risk Assessment of Connection Loss Fallback Strategies
eess.SY cs.SY
As part of the overall goal of driverless road vehicles, remote driving is a major emerging field of research of its own. Current remote driving concepts for public road traffic often establish a fallback strategy of immediate braking to a standstill in the event of a connection loss. This may seem like the most logical option when human control of the vehicle is lost. However, our simulation results from hundreds of scenarios based on naturalistic traffic scenes indicate high collision rates for any immediate substantial deceleration to a standstill in urban settings. We show that such a fallback strategy can result in a SOTIF relevant hazard, making it questionable whether such a design decision can be considered acceptable. Therefore, from a safety perspective, we would call this problem a safety blind spot, as safety analyses in this regard seem to be very rare. In this article, we first present a simulation on a naturalistic dataset that shows a high probability of collision in the described case. Second, we discuss the severity of the resulting potential rear-end collisions and provide an even more severe example by including a large commercial vehicle in the potential collision.
2502.10248
Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model
cs.CV cs.CL
We present Step-Video-T2V, a state-of-the-art text-to-video pre-trained model with 30B parameters and the ability to generate videos up to 204 frames in length. A deep compression Variational Autoencoder, Video-VAE, is designed for video generation tasks, achieving 16x16 spatial and 8x temporal compression ratios, while maintaining exceptional video reconstruction quality. User prompts are encoded using two bilingual text encoders to handle both English and Chinese. A DiT with 3D full attention is trained using Flow Matching and is employed to denoise input noise into latent frames. A video-based DPO approach, Video-DPO, is applied to reduce artifacts and improve the visual quality of the generated videos. We also detail our training strategies and share key observations and insights. Step-Video-T2V's performance is evaluated on a novel video generation benchmark, Step-Video-T2V-Eval, demonstrating its state-of-the-art text-to-video quality when compared with both open-source and commercial engines. Additionally, we discuss the limitations of current diffusion-based model paradigm and outline future directions for video foundation models. We make both Step-Video-T2V and Step-Video-T2V-Eval available at https://github.com/stepfun-ai/Step-Video-T2V. The online version can be accessed from https://yuewen.cn/videos as well. Our goal is to accelerate the innovation of video foundation models and empower video content creators.
2502.10250
VisCon-100K: Leveraging Contextual Web Data for Fine-tuning Vision Language Models
cs.CL cs.CV
Vision-language models (VLMs) excel in various visual benchmarks but are often constrained by the lack of high-quality visual fine-tuning data. To address this challenge, we introduce VisCon-100K, a novel dataset derived from interleaved image-text web documents. Our approach transforms 45K web documents from the OBELICS dataset into 100K image conversation samples. We utilize GPT-4V to generate image-contextual captions and OpenChat 3.5 model to convert these captions into diverse free-form and multiple-choice question-answer pairs. Integrating this dataset for fine-tuning considerably enhances VLM performance across multiple benchmarks. Unlike methods that focus solely on fine-grained visual content, our approach leverages accompanying web context, yielding superior results. We also discover that a `leaky modality mix,' where conversation samples contain questions answerable from both the image and its contextual caption, outperforms non-leaky combinations of captions and Q\&A pairs. VisCon-100k dataset shows strong performance with two popular VLM approaches: text-only large language model (LLM) aligned with a vision encoder using image captions data (ShareGPT4V-7b) and multimodally pretrained LLM (IDEFICS2-8b) using interleaved image-text data. In addition to releasing the VisCon-100K dataset, we provide a contextual captioner trained on this dataset, facilitating scalable fine-tuning data generation for future research and open-source applications. Using the same pipeline, but substituting our trained contextual captioner for GPT-4V, we also release the larger VisCon-1M dataset.
2502.10258
PromptArtisan: Multi-instruction Image Editing in Single Pass with Complete Attention Control
cs.CV cs.HC
We present PromptArtisan, a groundbreaking approach to multi-instruction image editing that achieves remarkable results in a single pass, eliminating the need for time-consuming iterative refinement. Our method empowers users to provide multiple editing instructions, each associated with a specific mask within the image. This flexibility allows for complex edits involving mask intersections or overlaps, enabling the realization of intricate and nuanced image transformations. PromptArtisan leverages a pre-trained InstructPix2Pix model in conjunction with a novel Complete Attention Control Mechanism (CACM). This mechanism ensures precise adherence to user instructions, granting fine-grained control over the editing process. Furthermore, our approach is zero-shot, requiring no additional training, and boasts improved processing complexity compared to traditional iterative methods. By seamlessly integrating multi-instruction capabilities, single-pass efficiency, and complete attention control, PromptArtisan unlocks new possibilities for creative and efficient image editing workflows, catering to both novice and expert users alike.
2502.10259
MITO: Enabling Non-Line-of-Sight Perception using Millimeter-waves through Real-World Datasets and Simulation Tools
cs.CV
We present MITO, the first dataset of multi-spectral millimeter-wave (mmWave) images of everyday objects. Unlike visible light, mmWave signals can image through everyday occlusions (e.g., cardboard boxes, fabric, plastic). However, due to the dearth of publicly-available mmWave images and the interdisciplinary challenges in collecting and processing mmWave signals, it remains difficult today for computer vision researchers to develop mmWave-based non-line-of-sight perception algorithms and models. To overcome these challenges, we introduce a real-world dataset and open-source simulation tool for mmWave imaging. The dataset is acquired using a UR5 robotic arm with two mmWave radars operating at different frequencies and an RGB-D camera. Through a signal processing pipeline, we capture and create over 580 real-world 3D mmWave images from over 76 different objects in the YCB dataset, a standard dataset for robotics manipulation. We provide real-world mmWave images in line-of-sight and non-line-of-sight, as well as RGB-D images and ground truth segmentation masks. We also develop an open-source simulation tool that can be used to generate synthetic mmWave images for any 3D triangle mesh, which achieves a median F-Score of 94% when compared to real-world mmWave images. We show the usefulness of this dataset and simulation tool in multiple CV tasks in non-line-of-sight. First, we perform object segmentation for mmWave images using the segment anything model (SAM), and achieve a median precision and recall of 92.6% and 64%. Second, we train a classifier that can recognize objects in non-line-of-sight. It is trained on synthetic images and can classify real-world images with 85% accuracy. We believe MITO will be a valuable resource for computer vision researchers in developing non-line-of-sight perception, similar to how early camera-based datasets shaped the field.
2502.10263
Large Language Models and Synthetic Data for Monitoring Dataset Mentions in Research Papers
cs.CL cs.AI cs.CY cs.DB cs.LG
Tracking how data is mentioned and used in research papers provides critical insights for improving data discoverability, quality, and production. However, manually identifying and classifying dataset mentions across vast academic literature is resource-intensive and not scalable. This paper presents a machine learning framework that automates dataset mention detection across research domains by leveraging large language models (LLMs), synthetic data, and a two-stage fine-tuning process. We employ zero-shot extraction from research papers, an LLM-as-a-Judge for quality assessment, and a reasoning agent for refinement to generate a weakly supervised synthetic dataset. The Phi-3.5-mini instruct model is pre-fine-tuned on this dataset, followed by fine-tuning on a manually annotated subset. At inference, a ModernBERT-based classifier efficiently filters dataset mentions, reducing computational overhead while maintaining high recall. Evaluated on a held-out manually annotated sample, our fine-tuned model outperforms NuExtract-v1.5 and GLiNER-large-v2.1 in dataset extraction accuracy. Our results highlight how LLM-generated synthetic data can effectively address training data scarcity, improving generalization in low-resource settings. This framework offers a pathway toward scalable monitoring of dataset usage, enhancing transparency, and supporting researchers, funders, and policymakers in identifying data gaps and strengthening data accessibility for informed decision-making.
2502.10266
Are Large Language Models the future crowd workers of Linguistics?
cs.CL cs.AI
Data elicitation from human participants is one of the core data collection strategies used in empirical linguistic research. The amount of participants in such studies may vary considerably, ranging from a handful to crowdsourcing dimensions. Even if they provide resourceful extensive data, both of these settings come alongside many disadvantages, such as low control of participants' attention during task completion, precarious working conditions in crowdsourcing environments, and time-consuming experimental designs. For these reasons, this research aims to answer the question of whether Large Language Models (LLMs) may overcome those obstacles if included in empirical linguistic pipelines. Two reproduction case studies are conducted to gain clarity into this matter: Cruz (2023) and Lombard et al. (2021). The two forced elicitation tasks, originally designed for human participants, are reproduced in the proposed framework with the help of OpenAI's GPT-4o-mini model. Its performance with our zero-shot prompting baseline shows the effectiveness and high versatility of LLMs, that tend to outperform human informants in linguistic tasks. The findings of the second replication further highlight the need to explore additional prompting techniques, such as Chain-of-Thought (CoT) prompting, which, in a second follow-up experiment, demonstrates higher alignment to human performance on both critical and filler items. Given the limited scale of this study, it is worthwhile to further explore the performance of LLMs in empirical Linguistics and in other future applications in the humanities.