id
stringlengths
9
16
title
stringlengths
4
278
categories
stringlengths
5
104
abstract
stringlengths
6
4.09k
2501.03225
Automated Generation of Challenging Multiple-Choice Questions for Vision Language Model Evaluation
cs.CV cs.AI cs.CL cs.CY cs.LG
The rapid development of vision language models (VLMs) demands rigorous and reliable evaluation. However, current visual question answering (VQA) benchmarks often depend on open-ended questions, making accurate evaluation difficult due to the variability in natural language responses. To address this, we introduce AutoConverter, an agentic framework that automatically converts these open-ended questions into multiple-choice format, enabling objective evaluation while reducing the costly question creation process. Our experiments demonstrate that AutoConverter can generate correct and challenging multiple-choice questions, with VLMs demonstrating consistently similar or lower accuracy on these questions compared to human-created ones. Using AutoConverter, we construct VMCBench, a benchmark created by transforming 20 existing VQA datasets into a unified multiple-choice format, totaling 9,018 questions. We comprehensively evaluate 33 state-of-the-art VLMs on VMCBench, setting a new standard for scalable, consistent, and reproducible VLM evaluation.
2501.03226
BoostStep: Boosting mathematical capability of Large Language Models via improved single-step reasoning
cs.CL cs.AI cs.LG
Large language models (LLMs) have demonstrated impressive ability in solving complex mathematical problems with multi-step reasoning and can be further enhanced with well-designed in-context learning (ICL) examples. However, this potential is often constrained by two major challenges in ICL: granularity mismatch and irrelevant information. We observe that while LLMs excel at decomposing mathematical problems, they often struggle with reasoning errors in fine-grained steps. Moreover, ICL examples retrieved at the question level may omit critical steps or even mislead the model with irrelevant details. To address this issue, we propose BoostStep, a method that enhances reasoning accuracy through step-aligned ICL, a novel mechanism that carefully aligns retrieved reference steps with the corresponding reasoning steps. Additionally, BoostStep incorporates an effective "first-try" strategy to deliver exemplars highly relevant to the current state of reasoning. BoostStep is a flexible and powerful method that integrates seamlessly with chain-of-thought (CoT) and tree search algorithms, refining both candidate selection and decision-making. Empirical results show that BoostStep improves GPT-4o's CoT performance by 4.6% across mathematical benchmarks, significantly surpassing traditional few-shot learning's 1.2%. Moreover, it can achieve an additional 7.5\% gain combined with tree search. Surprisingly, it enhances state-of-the-art LLMs to solve challenging math problems using simpler examples. It improves DeepSeek-R1-671B's performance on AIME by 2.2%, leveraging simple examples only from the MATH dataset.
2501.03227
When Should Selfish Miners Double-Spend?
cs.CR cs.DC cs.DM cs.IT math.IT math.PR
Although, both double-spending and selfish-mining attacks have been extensively studied since the ``Bitcoin'' whitepaper of Nakamoto and the ``majority is not enough'' paper of Eyal and Sirer, there has been no rigorous stochastic analysis of an attack that combines the two, except for the complicated MDP models. In this paper, we first combine stubborn and selfish mining attacks, i.e., construct a strategy where the attacker acts stubborn until its private branch reaches a certain length and then switches to act selfish. We provide the optimal stubbornness for each parameter regime. Next, we provide the maximum stubbornness that is still more profitable than honest mining and argue a connection between the level of stubbornness and the $k$-confirmation rule. We show that, at each attack cycle, if the level of stubbornness is higher than $k$, there is a risk of double-spending which comes at no-cost to the adversary. The result can be seen as a guide for picking $k$ in the $k$-confirmation rule in a blockchain design. At each cycle, for a given stubbornness level, we rigorously formulate how great the risk of double-spending is. We provide the minimum double-spend value needed for an attack to be profitable in the regimes where the scheme is less profitable than honest mining. We further modify the attack in the stubborn regime in order to conceal the attack and increase the double-spending probability. Finally, we evaluate the results and provide the optimal and the maximum stubbornness levels for each parameter regime as well as the revenue. As a case study, with Bitcoin's $k=6$ block confirmation rule, we evaluate the revenue and double-spending risk of the attacks for each pool parameter.
2501.03228
LightGNN: Simple Graph Neural Network for Recommendation
cs.IR cs.AI cs.LG
Graph neural networks (GNNs) have demonstrated superior performance in collaborative recommendation through their ability to conduct high-order representation smoothing, effectively capturing structural information within users' interaction patterns. However, existing GNN paradigms face significant challenges in scalability and robustness when handling large-scale, noisy, and real-world datasets. To address these challenges, we present LightGNN, a lightweight and distillation-based GNN pruning framework designed to substantially reduce model complexity while preserving essential collaboration modeling capabilities. Our LightGNN framework introduces a computationally efficient pruning module that adaptively identifies and removes redundant edges and embedding entries for model compression. The framework is guided by a resource-friendly hierarchical knowledge distillation objective, whose intermediate layer augments the observed graph to maintain performance, particularly in high-rate compression scenarios. Extensive experiments on public datasets demonstrate LightGNN's effectiveness, significantly improving both computational efficiency and recommendation accuracy. Notably, LightGNN achieves an 80% reduction in edge count and 90% reduction in embedding entries while maintaining performance comparable to more complex state-of-the-art baselines. The implementation of our LightGNN framework is available at the github repository: https://github.com/HKUDS/LightGNN.
2501.03229
Gaussian Masked Autoencoders
cs.CV cs.AI
This paper explores Masked Autoencoders (MAE) with Gaussian Splatting. While reconstructive self-supervised learning frameworks such as MAE learns good semantic abstractions, it is not trained for explicit spatial awareness. Our approach, named Gaussian Masked Autoencoder, or GMAE, aims to learn semantic abstractions and spatial understanding jointly. Like MAE, it reconstructs the image end-to-end in the pixel space, but beyond MAE, it also introduces an intermediate, 3D Gaussian-based representation and renders images via splatting. We show that GMAE can enable various zero-shot learning capabilities of spatial understanding (e.g., figure-ground segmentation, image layering, edge detection, etc.) while preserving the high-level semantics of self-supervised representation quality from MAE. To our knowledge, we are the first to employ Gaussian primitives in an image representation learning framework beyond optimization-based single-scene reconstructions. We believe GMAE will inspire further research in this direction and contribute to developing next-generation techniques for modeling high-fidelity visual data. More details at https://brjathu.github.io/gmae
2501.03230
Video-of-Thought: Step-by-Step Video Reasoning from Perception to Cognition
cs.AI cs.CV
Existing research of video understanding still struggles to achieve in-depth comprehension and reasoning in complex videos, primarily due to the under-exploration of two key bottlenecks: fine-grained spatial-temporal perceptive understanding and cognitive-level video scene comprehension. This paper bridges the gap by presenting a novel solution. We first introduce a novel video Multimodal Large Language Model (MLLM), MotionEpic, which achieves fine-grained pixel-level spatial-temporal video grounding by integrating video spatial-temporal scene graph (STSG) representation. Building upon MotionEpic, we then develop a Video-of-Thought (VoT) reasoning framework. VoT inherits the Chain-of-Thought (CoT) core, breaking down a complex task into simpler and manageable sub-problems, and addressing them step-by-step from a low-level pixel perception to high-level cognitive interpretation. Extensive experiments across various complex video QA benchmarks demonstrate that our overall framework strikingly boosts existing state-of-the-art. To our knowledge, this is the first attempt at successfully implementing the CoT technique for achieving human-level video reasoning, where we show great potential in extending it to a wider range of video understanding scenarios. Project is open at https://haofei.vip/VoT
2501.03235
Neural networks consisting of DNA
physics.bio-ph cond-mat.soft cs.AI cs.NE q-bio.BM q-bio.MN
Neural networks based on soft and biological matter constitute an interesting potential alternative to traditional implementations based on electric circuits. DNA is a particularly promising system in this context due its natural ability to store information. In recent years, researchers have started to construct neural networks that are based on DNA. In this chapter, I provide a very basic introduction to the concept of DNA neural networks, aiming at an audience that is not familiar with biochemistry.
2501.03246
Bridging Auditory Perception and Language Comprehension through MEG-Driven Encoding Models
q-bio.NC cs.CL cs.LG cs.SD eess.AS eess.SP
Understanding the neural mechanisms behind auditory and linguistic processing is key to advancing cognitive neuroscience. In this study, we use Magnetoencephalography (MEG) data to analyze brain responses to spoken language stimuli. We develop two distinct encoding models: an audio-to-MEG encoder, which uses time-frequency decompositions (TFD) and wav2vec2 latent space representations, and a text-to-MEG encoder, which leverages CLIP and GPT-2 embeddings. Both models successfully predict neural activity, demonstrating significant correlations between estimated and observed MEG signals. However, the text-to-MEG model outperforms the audio-based model, achieving higher Pearson Correlation (PC) score. Spatially, we identify that auditory-based embeddings (TFD and wav2vec2) predominantly activate lateral temporal regions, which are responsible for primary auditory processing and the integration of auditory signals. In contrast, textual embeddings (CLIP and GPT-2) primarily engage the frontal cortex, particularly Broca's area, which is associated with higher-order language processing, including semantic integration and language production, especially in the 8-30 Hz frequency range. The strong involvement of these regions suggests that auditory stimuli are processed through more direct sensory pathways, while linguistic information is encoded via networks that integrate meaning and cognitive control. Our results reveal distinct neural pathways for auditory and linguistic information processing, with higher encoding accuracy for text representations in the frontal regions. These insights refine our understanding of the brain's functional architecture in processing auditory and textual information, offering quantitative advancements in the modelling of neural responses to complex language stimuli.
2501.03250
Machine Learning and Deep Learning Techniques used in Cybersecurity and Digital Forensics: a Review
cs.CR cs.AI
In the paced realms of cybersecurity and digital forensics machine learning (ML) and deep learning (DL) have emerged as game changing technologies that introduce methods to identify stop and analyze cyber risks. This review presents an overview of the ML and DL approaches used in these fields showcasing their advantages drawbacks and possibilities. It covers a range of AI techniques used in spotting intrusions in systems and classifying malware to prevent cybersecurity attacks, detect anomalies and enhance resilience. This study concludes by highlighting areas where further research is needed and suggesting ways to create transparent and scalable ML and DL solutions that are suited to the evolving landscape of cybersecurity and digital forensics.
2501.03254
Advanced Displacement Magnitude Prediction in Multi-Material Architected Lattice Structure Beams Using Physics Informed Neural Network Architecture
cs.AI cond-mat.mtrl-sci cs.CE cs.LG cs.NE
This paper proposes an innovative method for predicting deformation in architected lattice structures that combines Physics-Informed Neural Networks (PINNs) with finite element analysis. A thorough study was carried out on FCC-based lattice beams utilizing five different materials (Structural Steel, AA6061, AA7075, Ti6Al4V, and Inconel 718) under varied edge loads (1000-10000 N). The PINN model blends data-driven learning with physics-based limitations via a proprietary loss function, resulting in much higher prediction accuracy than linear regression. PINN outperforms linear regression, achieving greater R-square (0.7923 vs 0.5686) and lower error metrics (MSE: 0.00017417 vs 0.00036187). Among the materials examined, AA6061 had the highest displacement sensitivity (0.1014 mm at maximum load), while Inconel718 had better structural stability.
2501.03256
AI-ANNE: (A) (N)eural (N)et for (E)xploration: Transferring Deep Learning Models onto Microcontrollers and Embedded Systems
cs.LG cs.AI
This working paper explores the integration of neural networks onto resource-constrained embedded systems like a Raspberry Pi Pico / Raspberry Pi Pico 2. A TinyML aproach transfers neural networks directly on these microcontrollers, enabling real-time, low-latency, and energy-efficient inference while maintaining data privacy. Therefore, AI-ANNE: (A) (N)eural (N)et for (E)xploration will be presented, which facilitates the transfer of pre-trained models from high-performance platforms like TensorFlow and Keras onto microcontrollers, using a lightweight programming language like MicroPython. This approach demonstrates how neural network architectures, such as neurons, layers, density and activation functions can be implemented in MicroPython in order to deal with the computational limitations of embedded systems. Based on the Raspberry Pi Pico / Raspberry Pi Pico 2, two different neural networks on microcontrollers are presented for an example of data classification. As an further application example, such a microcontroller can be used for condition monitoring, where immediate corrective measures are triggered on the basis of sensor data. Overall, this working paper presents a very easy-to-implement way of using neural networks on energy-efficient devices such as microcontrollers. This makes AI-ANNE: (A) (N)eural (N)et for (E)xploration not only suited for practical use, but also as an educational tool with clear insights into how neural networks operate.
2501.03257
Breaking Through the Spike: Spike Window Decoding for Accelerated and Precise Automatic Speech Recognition
eess.AS cs.AI cs.CL cs.SD
Recently, end-to-end automatic speech recognition has become the mainstream approach in both industry and academia. To optimize system performance in specific scenarios, the Weighted Finite-State Transducer (WFST) is extensively used to integrate acoustic and language models, leveraging its capacity to implicitly fuse language models within static graphs, thereby ensuring robust recognition while also facilitating rapid error correction. However, WFST necessitates a frame-by-frame search of CTC posterior probabilities through autoregression, which significantly hampers inference speed. In this work, we thoroughly investigate the spike property of CTC outputs and further propose the conjecture that adjacent frames to non-blank spikes carry semantic information beneficial to the model. Building on this, we propose the Spike Window Decoding algorithm, which greatly improves the inference speed by making the number of frames decoded in WFST linearly related to the number of spiking frames in the CTC output, while guaranteeing the recognition performance. Our method achieves SOTA recognition accuracy with significantly accelerates decoding speed, proven across both AISHELL-1 and large-scale In-House datasets, establishing a pioneering approach for integrating CTC output with WFST.
2501.03259
Toward Inclusive Educational AI: Auditing Frontier LLMs through a Multiplexity Lens
cs.CL cs.AI cs.CY cs.LG cs.MA
As large language models (LLMs) like GPT-4 and Llama 3 become integral to educational contexts, concerns are mounting over the cultural biases, power imbalances, and ethical limitations embedded within these technologies. Though generative AI tools aim to enhance learning experiences, they often reflect values rooted in Western, Educated, Industrialized, Rich, and Democratic (WEIRD) cultural paradigms, potentially sidelining diverse global perspectives. This paper proposes a framework to assess and mitigate cultural bias within LLMs through the lens of applied multiplexity. Multiplexity, inspired by Senturk et al. and rooted in Islamic and other wisdom traditions, emphasizes the coexistence of diverse cultural viewpoints, supporting a multi-layered epistemology that integrates both empirical sciences and normative values. Our analysis reveals that LLMs frequently exhibit cultural polarization, with biases appearing in both overt responses and subtle contextual cues. To address inherent biases and incorporate multiplexity in LLMs, we propose two strategies: \textit{Contextually-Implemented Multiplex LLMs}, which embed multiplex principles directly into the system prompt, influencing LLM outputs at a foundational level and independent of individual prompts, and \textit{Multi-Agent System (MAS)-Implemented Multiplex LLMs}, where multiple LLM agents, each representing distinct cultural viewpoints, collaboratively generate a balanced, synthesized response. Our findings demonstrate that as mitigation strategies evolve from contextual prompting to MAS-implementation, cultural inclusivity markedly improves, evidenced by a significant rise in the Perspectives Distribution Score (PDS) and a PDS Entropy increase from 3.25\% at baseline to 98\% with the MAS-Implemented Multiplex LLMs. Sentiment analysis further shows a shift towards positive sentiment across cultures,...
2501.03261
Navigation Variable-based Multi-objective Particle Swarm Optimization for UAV Path Planning with Kinematic Constraints
cs.RO cs.AI cs.NE
Path planning is essential for unmanned aerial vehicles (UAVs) as it determines the path that the UAV needs to follow to complete a task. This work addresses this problem by introducing a new algorithm called navigation variable-based multi-objective particle swarm optimization (NMOPSO). It first models path planning as an optimization problem via the definition of a set of objective functions that include optimality and safety requirements for UAV operation. The NMOPSO is then used to minimize those functions through Pareto optimal solutions. The algorithm features a new path representation based on navigation variables to include kinematic constraints and exploit the maneuverable characteristics of the UAV. It also includes an adaptive mutation mechanism to enhance the diversity of the swarm for better solutions. Comparisons with various algorithms have been carried out to benchmark the proposed approach. The results indicate that the NMOPSO performs better than not only other particle swarm optimization variants but also other state-of-the-art multi-objective and metaheuristic optimization algorithms. Experiments have also been conducted with real UAVs to confirm the validity of the approach for practical flights. The source code of the algorithm is available at https://github.com/ngandng/NMOPSO.
2501.03262
REINFORCE++: A Simple and Efficient Approach for Aligning Large Language Models
cs.CL cs.LG
Reinforcement Learning from Human Feedback (RLHF) has emerged as a critical approach for aligning large language models with human preferences, witnessing rapid algorithmic evolution through methods such as Proximal Policy Optimization (PPO), Direct Preference Optimization (DPO), REINFORCE Leave One-Out (RLOO), ReMax, and Group Relative Policy Optimization (GRPO). We present REINFORCE++, an enhanced variant of the classical REINFORCE algorithm that incorporates key optimization techniques from PPO while eliminating the need for a critic network. REINFORCE++ achieves three primary objectives: (1) simplicity (2) enhanced training stability, and (3) reduced computational overhead. Through extensive empirical evaluation, we demonstrate that REINFORCE++ exhibits superior stability compared to GRPO and achieves greater computational efficiency than PPO while maintaining comparable performance. The implementation is available at \url{https://github.com/OpenRLHF/OpenRLHF}.
2501.03264
Bridge the Inference Gaps of Neural Processes via Expectation Maximization
cs.LG cs.AI cs.NE
The neural process (NP) is a family of computationally efficient models for learning distributions over functions. However, it suffers from under-fitting and shows suboptimal performance in practice. Researchers have primarily focused on incorporating diverse structural inductive biases, \textit{e.g.} attention or convolution, in modeling. The topic of inference suboptimality and an analysis of the NP from the optimization objective perspective has hardly been studied in earlier work. To fix this issue, we propose a surrogate objective of the target log-likelihood of the meta dataset within the expectation maximization framework. The resulting model, referred to as the Self-normalized Importance weighted Neural Process (SI-NP), can learn a more accurate functional prior and has an improvement guarantee concerning the target log-likelihood. Experimental results show the competitive performance of SI-NP over other NPs objectives and illustrate that structural inductive biases, such as attention modules, can also augment our method to achieve SOTA performance. Our code is available at \url{https://github.com/hhq123gogogo/SI_NPs}.
2501.03265
Optimizing Edge AI: A Comprehensive Survey on Data, Model, and System Strategies
cs.LG cs.AI
The emergence of 5G and edge computing hardware has brought about a significant shift in artificial intelligence, with edge AI becoming a crucial technology for enabling intelligent applications. With the growing amount of data generated and stored on edge devices, deploying AI models for local processing and inference has become increasingly necessary. However, deploying state-of-the-art AI models on resource-constrained edge devices faces significant challenges that must be addressed. This paper presents an optimization triad for efficient and reliable edge AI deployment, including data, model, and system optimization. First, we discuss optimizing data through data cleaning, compression, and augmentation to make it more suitable for edge deployment. Second, we explore model design and compression methods at the model level, such as pruning, quantization, and knowledge distillation. Finally, we introduce system optimization techniques like framework support and hardware acceleration to accelerate edge AI workflows. Based on an in-depth analysis of various application scenarios and deployment challenges of edge AI, this paper proposes an optimization paradigm based on the data-model-system triad to enable a whole set of solutions to effectively transfer ML models, which are initially trained in the cloud, to various edge devices for supporting multiple scenarios.
2501.03266
LLM Content Moderation and User Satisfaction: Evidence from Response Refusals in Chatbot Arena
cs.CL cs.AI cs.CY cs.HC cs.SI
LLM safety and ethical alignment are widely discussed, but the impact of content moderation on user satisfaction remains underexplored. To address this, we analyze nearly 50,000 Chatbot Arena response-pairs using a novel fine-tuned RoBERTa model, that we trained on hand-labeled data to disentangle refusals due to ethical concerns from other refusals due to technical disabilities or lack of information. Our findings reveal a significant refusal penalty on content moderation, with users choosing ethical-based refusals roughly one-fourth as often as their preferred LLM response compared to standard responses. However, the context and phrasing play critical roles: refusals on highly sensitive prompts, such as illegal content, achieve higher win rates than less sensitive ethical concerns, and longer responses closely aligned with the prompt perform better. These results emphasize the need for nuanced moderation strategies that balance ethical safeguards with user satisfaction. Moreover, we find that the refusal penalty is notably lower in evaluations using the LLM-as-a-Judge method, highlighting discrepancies between user and automated assessments.
2501.03268
Heterogeneous Graph Pre-training Based Model for Secure and Efficient Prediction of Default Risk Propagation among Bond Issuers
cs.LG cs.AI
Efficient prediction of default risk for bond-issuing enterprises is pivotal for maintaining stability and fostering growth in the bond market. Conventional methods usually rely solely on an enterprise's internal data for risk assessment. In contrast, graph-based techniques leverage interconnected corporate information to enhance default risk identification for targeted bond issuers. Traditional graph techniques such as label propagation algorithm or deepwalk fail to effectively integrate a enterprise's inherent attribute information with its topological network data. Additionally, due to data scarcity and security privacy concerns between enterprises, end-to-end graph neural network (GNN) algorithms may struggle in delivering satisfactory performance for target tasks. To address these challenges, we present a novel two-stage model. In the first stage, we employ an innovative Masked Autoencoders for Heterogeneous Graph (HGMAE) to pre-train on a vast enterprise knowledge graph. Subsequently, in the second stage, a specialized classifier model is trained to predict default risk propagation probabilities. The classifier leverages concatenated feature vectors derived from the pre-trained encoder with the enterprise's task-specific feature vectors. Through the two-stage training approach, our model not only boosts the importance of unique bond characteristics for specific default prediction tasks, but also securely and efficiently leverage the global information pre-trained from other enterprises. Experimental results demonstrate that our proposed model outperforms existing approaches in predicting default risk for bond issuers.
2501.03271
DPO Kernels: A Semantically-Aware, Kernel-Enhanced, and Divergence-Rich Paradigm for Direct Preference Optimization
cs.LG cs.AI cs.CL
The rapid rise of large language models (LLMs) has unlocked many applications but also underscores the challenge of aligning them with diverse values and preferences. Direct Preference Optimization (DPO) is central to alignment but constrained by fixed divergences and limited feature transformations. We propose DPO-Kernels, which integrates kernel methods to address these issues through four key contributions: (i) Kernelized Representations with polynomial, RBF, Mahalanobis, and spectral kernels for richer transformations, plus a hybrid loss combining embedding-based and probability-based objectives; (ii) Divergence Alternatives (Jensen-Shannon, Hellinger, Renyi, Bhattacharyya, Wasserstein, and f-divergences) for greater stability; (iii) Data-Driven Selection metrics that automatically choose the best kernel-divergence pair; and (iv) a Hierarchical Mixture of Kernels for both local precision and global modeling. Evaluations on 12 datasets demonstrate state-of-the-art performance in factuality, safety, reasoning, and instruction following. Grounded in Heavy-Tailed Self-Regularization, DPO-Kernels maintains robust generalization for LLMs, offering a comprehensive resource for further alignment research.
2501.03272
Backdoor Token Unlearning: Exposing and Defending Backdoors in Pretrained Language Models
cs.CR cs.AI cs.CL
Supervised fine-tuning has become the predominant method for adapting large pretrained models to downstream tasks. However, recent studies have revealed that these models are vulnerable to backdoor attacks, where even a small number of malicious samples can successfully embed backdoor triggers into the model. While most existing defense methods focus on post-training backdoor defense, efficiently defending against backdoor attacks during training phase remains largely unexplored. To address this gap, we propose a novel defense method called Backdoor Token Unlearning (BTU), which proactively detects and neutralizes trigger tokens during the training stage. Our work is based on two key findings: 1) backdoor learning causes distinctive differences between backdoor token parameters and clean token parameters in word embedding layers, and 2) the success of backdoor attacks heavily depends on backdoor token parameters. The BTU defense leverages these properties to identify aberrant embedding parameters and subsequently removes backdoor behaviors using a fine-grained unlearning technique. Extensive evaluations across three datasets and four types of backdoor attacks demonstrate that BTU effectively defends against these threats while preserving the model's performance on primary tasks. Our code is available at https://github.com/XDJPH/BTU.
2501.03273
Strategic Fusion Optimizes Transformer Compression
cs.LG cs.AI cs.CL
This study investigates transformer model compression by systematically pruning its layers. We evaluated 14 pruning strategies across nine diverse datasets, including 12 strategies based on different signals obtained from layer activations, mutual information, gradients, weights, and attention. To address the limitations of single-signal strategies, we introduced two fusion strategies, linear regression and random forest, which combine individual strategies (i.e., strategic fusion), for more informed pruning decisions. Additionally, we applied knowledge distillation to mitigate any accuracy loss during layer pruning. Our results reveal that random forest strategic fusion outperforms individual strategies in seven out of nine datasets and achieves near-optimal performance in the other two. The distilled random forest surpasses the original accuracy in six datasets and mitigates accuracy drops in the remaining three. Knowledge distillation also improves the accuracy-to-size ratio by an average factor of 18.84 across all datasets. Supported by mathematical foundations and biological analogies, our findings suggest that strategically combining multiple signals can lead to efficient, high-performing transformer models for resource-constrained applications.
2501.03276
ComMer: a Framework for Compressing and Merging User Data for Personalization
cs.CL cs.AI cs.IR cs.LG
Large Language Models (LLMs) excel at a wide range of tasks, but adapting them to new data, particularly for personalized applications, poses significant challenges due to resource and computational constraints. Existing methods either rely on exposing fresh data to the model through the prompt, which is limited by context size and computationally expensive at inference time, or fine-tuning, which incurs substantial training and update costs. In this paper, we introduce ComMer - Compress and Merge - a novel framework that efficiently personalizes LLMs by compressing users' documents into compact representations, which are then merged and fed into a frozen LLM. We evaluate ComMer on two types of personalization tasks - personalized skill learning, using the tweet paraphrasing dataset and the personalized news headline generation dataset from the LaMP benchmark, and knowledge-intensive, using the PerLTQA dataset. Our experiments demonstrate that in constrained inference budget scenarios ComMer achieves superior quality in skill learning tasks, while highlighting limitations in knowledge-intensive settings due to the loss of detailed information. These results offer insights into trade-offs and potential optimizations in multi-document compression for personalization.
2501.03277
HonkaiChat: Companions from Anime that feel alive!
cs.CL
Modern conversational agents, including anime-themed chatbots, are frequently reactive and personality-driven but fail to capture the dynamic nature of human interactions. We propose an event-driven dialogue framework to address these limitations by embedding dynamic events in conversation prompts and fine-tuning models on character-specific data. Evaluations on GPT-4 and comparisons with industry-leading baselines demonstrate that event-driven prompts significantly improve conversational engagement and naturalness while reducing hallucinations. This paper explores the application of this approach in creating lifelike chatbot interactions within the context of Honkai: Star Rail, showcasing the potential for dynamic event-based systems to transform role-playing and interactive dialogue.
2501.03278
DenseGNN: universal and scalable deeper graph neural networks for high-performance property prediction in crystals and molecules
cond-mat.mtrl-sci cs.LG
Generative models generate vast numbers of hypothetical materials, necessitating fast, accurate models for property prediction. Graph Neural Networks (GNNs) excel in this domain but face challenges like high training costs, domain adaptation issues, and over-smoothing. We introduce DenseGNN, which employs Dense Connectivity Network (DCN), Hierarchical Node-Edge-Graph Residual Networks (HRN), and Local Structure Order Parameters Embedding (LOPE) to address these challenges. DenseGNN achieves state-of-the-art performance on datasets such as JARVIS-DFT, Materials Project, and QM9, improving the performance of models like GIN, Schnet, and Hamnet on materials datasets. By optimizing atomic embeddings and reducing computational costs, DenseGNN enables deeper architectures and surpasses other GNNs in crystal structure distinction, approaching X-ray diffraction method accuracy. This advances materials discovery and design.
2501.03279
Revolutionizing Encrypted Traffic Classification with MH-Net: A Multi-View Heterogeneous Graph Model
cs.CR cs.AI cs.LG
With the growing significance of network security, the classification of encrypted traffic has emerged as an urgent challenge. Traditional byte-based traffic analysis methods are constrained by the rigid granularity of information and fail to fully exploit the diverse correlations between bytes. To address these limitations, this paper introduces MH-Net, a novel approach for classifying network traffic that leverages multi-view heterogeneous traffic graphs to model the intricate relationships between traffic bytes. The essence of MH-Net lies in aggregating varying numbers of traffic bits into multiple types of traffic units, thereby constructing multi-view traffic graphs with diverse information granularities. By accounting for different types of byte correlations, such as header-payload relationships, MH-Net further endows the traffic graph with heterogeneity, significantly enhancing model performance. Notably, we employ contrastive learning in a multi-task manner to strengthen the robustness of the learned traffic unit representations. Experiments conducted on the ISCX and CIC-IoT datasets for both the packet-level and flow-level traffic classification tasks demonstrate that MH-Net achieves the best overall performance compared to dozens of SOTA methods.
2501.03282
From Aleatoric to Epistemic: Exploring Uncertainty Quantification Techniques in Artificial Intelligence
cs.AI cs.LG
Uncertainty quantification (UQ) is a critical aspect of artificial intelligence (AI) systems, particularly in high-risk domains such as healthcare, autonomous systems, and financial technology, where decision-making processes must account for uncertainty. This review explores the evolution of uncertainty quantification techniques in AI, distinguishing between aleatoric and epistemic uncertainties, and discusses the mathematical foundations and methods used to quantify these uncertainties. We provide an overview of advanced techniques, including probabilistic methods, ensemble learning, sampling-based approaches, and generative models, while also highlighting hybrid approaches that integrate domain-specific knowledge. Furthermore, we examine the diverse applications of UQ across various fields, emphasizing its impact on decision-making, predictive accuracy, and system robustness. The review also addresses key challenges such as scalability, efficiency, and integration with explainable AI, and outlines future directions for research in this rapidly developing area. Through this comprehensive survey, we aim to provide a deeper understanding of UQ's role in enhancing the reliability, safety, and trustworthiness of AI systems.
2501.03284
Sensorformer: Cross-patch attention with global-patch compression is effective for high-dimensional multivariate time series forecasting
cs.LG
Among the existing Transformer-based multivariate time series forecasting methods, iTransformer, which treats each variable sequence as a token and only explicitly extracts cross-variable dependencies, and PatchTST, which adopts a channel-independent strategy and only explicitly extracts cross-time dependencies, both significantly outperform most Channel-Dependent Transformer that simultaneously extract cross-time and cross-variable dependencies. This indicates that existing Transformer-based multivariate time series forecasting methods still struggle to effectively fuse these two types of information. We attribute this issue to the dynamic time lags in the causal relationships between different variables. Therefore, we propose a new multivariate time series forecasting Transformer, Sensorformer, which first compresses the global patch information and then simultaneously extracts cross-variable and cross-time dependencies from the compressed representations. Sensorformer can effectively capture the correct inter-variable correlations and causal relationships, even in the presence of dynamic causal lags between variables, while also reducing the computational complexity of pure cross-patch self-attention from $O(D^2 \cdot Patch\_num^2 \cdot d\_model)$ to $O(D^2 \cdot Patch\_num \cdot d\_model)$. Extensive comparative and ablation experiments on 9 mainstream real-world multivariate time series forecasting datasets demonstrate the superiority of Sensorformer. The implementation of Sensorformer, following the style of the Time-series-library and scripts for reproducing the main results, is publicly available at https://github.com/BigYellowTiger/Sensorformer
2501.03286
Inverse Design of Optimal Stern Shape with Convolutional Neural Network-based Pressure Distribution
cs.LG physics.flu-dyn
Hull form designing is an iterative process wherein the performance of the hull form needs to be checked via computational fluid dynamics calculations or model experiments. The stern shape has to undergo a process wherein the hull form variations from the pressure distribution analysis results are repeated until the resistance and propulsion efficiency meet the design requirements. In this study, the designer designed a pressure distribution that meets the design requirements; this paper proposes an inverse design algorithm that estimates the stern shape using deep learning. A convolutional neural network was used to extract the features of the pressure distribution expressed as a contour, whereas a multi-task learning model was used to estimate various sections of the stern shape. We estimated the stern shape indirectly by estimating the control point of the B-spline and comparing the actual and converted offsets for each section; the performance was verified, and an inverse design is proposed herein
2501.03287
OpenLKA: an open dataset of lane keeping assist from market autonomous vehicles
cs.RO cs.CV cs.LG
The Lane Keeping Assist (LKA) system has become a standard feature in recent car models. While marketed as providing auto-steering capabilities, the system's operational characteristics and safety performance remain underexplored, primarily due to a lack of real-world testing and comprehensive data. To fill this gap, we extensively tested mainstream LKA systems from leading U.S. automakers in Tampa, Florida. Using an innovative method, we collected a comprehensive dataset that includes full Controller Area Network (CAN) messages with LKA attributes, as well as video, perception, and lateral trajectory data from a high-quality front-facing camera equipped with advanced vision detection and trajectory planning algorithms. Our tests spanned diverse, challenging conditions, including complex road geometry, adverse weather, degraded lane markings, and their combinations. A vision language model (VLM) further annotated the videos to capture weather, lighting, and traffic features. Based on this dataset, we present an empirical overview of LKA's operational features and safety performance. Key findings indicate: (i) LKA is vulnerable to faint markings and low pavement contrast; (ii) it struggles in lane transitions (merges, diverges, intersections), often causing unintended departures or disengagements; (iii) steering torque limitations lead to frequent deviations on sharp turns, posing safety risks; and (iv) LKA systems consistently maintain rigid lane-centering, lacking adaptability on tight curves or near large vehicles such as trucks. We conclude by demonstrating how this dataset can guide both infrastructure planning and self-driving technology. In view of LKA's limitations, we recommend improvements in road geometry and pavement maintenance. Additionally, we illustrate how the dataset supports the development of human-like LKA systems via VLM fine-tuning and Chain of Thought reasoning.
2501.03288
CodeVision: Detecting LLM-Generated Code Using 2D Token Probability Maps and Vision Models
cs.SE cs.AI
The rise of large language models (LLMs) like ChatGPT has significantly improved automated code generation, enhancing software development efficiency. However, this introduces challenges in academia, particularly in distinguishing between human-written and LLM-generated code, which complicates issues of academic integrity. Existing detection methods, such as pre-trained models and watermarking, face limitations in adaptability and computational efficiency. In this paper, we propose a novel detection method using 2D token probability maps combined with vision models, preserving spatial code structures such as indentation and brackets. By transforming code into log probability matrices and applying vision models like Vision Transformers (ViT) and ResNet, we capture both content and structure for more accurate detection. Our method shows robustness across multiple programming languages and improves upon traditional detectors, offering a scalable and computationally efficient solution for identifying LLM-generated code.
2501.03289
Adaptive Pruning of Pretrained Transformer via Differential Inclusions
cs.LG
Large transformers have demonstrated remarkable success, making it necessary to compress these models to reduce inference costs while preserving their perfor-mance. Current compression algorithms prune transformers at fixed compression ratios, requiring a unique pruning process for each ratio, which results in high computational costs. In contrast, we propose pruning of pretrained transformers at any desired ratio within a single pruning stage, based on a differential inclusion for a mask parameter. This dynamic can generate the whole regularization solution path of the mask parameter, whose support set identifies the network structure. Therefore, the solution path identifies a Transformer weight family with various sparsity levels, offering greater flexibility and customization. In this paper, we introduce such an effective pruning method, termed SPP (Solution Path Pruning). To achieve effective pruning, we segment the transformers into paired modules, including query-key pairs, value-projection pairs, and sequential linear layers, and apply low-rank compression to these pairs, maintaining the output structure while enabling structural compression within the inner states. Extensive experiments conducted on various well-known transformer backbones have demonstrated the efficacy of SPP.
2501.03290
A Decision-Based Heterogenous Graph Attention Network for Multi-Class Fake News Detection
cs.LG cs.AI cs.SI
A promising tool for addressing fake news detection is Graph Neural Networks (GNNs). However, most existing GNN-based methods rely on binary classification, categorizing news as either real or fake. Additionally, traditional GNN models use a static neighborhood for each node, making them susceptible to issues like over-squashing. In this paper, we introduce a novel model named Decision-based Heterogeneous Graph Attention Network (DHGAT) for fake news detection in a semi-supervised setting. DHGAT effectively addresses the limitations of traditional GNNs by dynamically optimizing and selecting the neighborhood type for each node in every layer. It represents news data as a heterogeneous graph where nodes (news items) are connected by various types of edges. The architecture of DHGAT consists of a decision network that determines the optimal neighborhood type and a representation network that updates node embeddings based on this selection. As a result, each node learns an optimal and task-specific computational graph, enhancing both the accuracy and efficiency of the fake news detection process. We evaluate DHGAT on the LIAR dataset, a large and challenging dataset for multi-class fake news detection, which includes news items categorized into six classes. Our results demonstrate that DHGAT outperforms existing methods, improving accuracy by approximately 4% and showing robustness with limited labeled data.
2501.03291
ADePT: Adaptive Decomposed Prompt Tuning for Parameter-Efficient Fine-tuning
cs.CL
Prompt Tuning (PT) enables the adaptation of Pre-trained Large Language Models (PLMs) to downstream tasks by optimizing a small amount of soft virtual tokens, which are prepended to the input token embeddings. Recently, Decomposed Prompt Tuning (DePT) has demonstrated superior adaptation capabilities by decomposing the soft prompt into a shorter soft prompt and a pair of low-rank matrices. The product of the pair of low-rank matrices is added to the input token embeddings to offset them. Additionally, DePT achieves faster inference compared to PT due to the shorter soft prompt. However, in this paper, we find that the position-based token embedding offsets of DePT restricts its ability to generalize across diverse model inputs, and that the shared embedding offsets across many token embeddings result in sub-optimization. To tackle these issues, we introduce \textbf{A}daptive \textbf{De}composed \textbf{P}rompt \textbf{T}uning (ADePT), which is composed of a short soft prompt and a shallow token-shared feed-forward neural network. ADePT utilizes the token-shared feed-forward neural network to learn the embedding offsets for each token, enabling adaptive embedding offsets that vary according to the model input and better optimization of token embedding offsets. This enables ADePT to achieve superior adaptation performance without requiring more inference time or additional trainable parameters compared to vanilla PT and its variants. In comprehensive experiments across 23 natural language processing (NLP) tasks and 4 typical PLMs of different scales, we show that ADePT consistently surpasses the leading parameter-efficient fine-tuning (PEFT) methods, and even outperforms the full fine-tuning baseline in certain scenarios. Code is available at \url{https://github.com/HungerPWAY/ADePT}.
2501.03292
Multi-Modal One-Shot Federated Ensemble Learning for Medical Data with Vision Large Language Model
cs.LG cs.AI
Federated learning (FL) has attracted considerable interest in the medical domain due to its capacity to facilitate collaborative model training while maintaining data privacy. However, conventional FL methods typically necessitate multiple communication rounds, leading to significant communication overhead and delays, especially in environments with limited bandwidth. One-shot federated learning addresses these issues by conducting model training and aggregation in a single communication round, thereby reducing communication costs while preserving privacy. Among these, one-shot federated ensemble learning combines independently trained client models using ensemble techniques such as voting, further boosting performance in non-IID data scenarios. On the other hand, existing machine learning methods in healthcare predominantly use unimodal data (e.g., medical images or textual reports), which restricts their diagnostic accuracy and comprehensiveness. Therefore, the integration of multi-modal data is proposed to address these shortcomings. In this paper, we introduce FedMME, an innovative one-shot multi-modal federated ensemble learning framework that utilizes multi-modal data for medical image analysis. Specifically, FedMME capitalizes on vision large language models to produce textual reports from medical images, employs a BERT model to extract textual features from these reports, and amalgamates these features with visual features to improve diagnostic accuracy. Experimental results show that our method demonstrated superior performance compared to existing one-shot federated learning methods in healthcare scenarios across four datasets with various data distributions. For instance, it surpasses existing one-shot federated learning approaches by more than 17.5% in accuracy on the RSNA dataset when applying a Dirichlet distribution with ($\alpha$ = 0.3).
2501.03295
A Soft Sensor Method with Uncertainty-Awareness and Self-Explanation Based on Large Language Models Enhanced by Domain Knowledge Retrieval
cs.LG cs.AI eess.SP
Data-driven soft sensors are crucial in predicting key performance indicators in industrial systems. However, current methods predominantly rely on the supervised learning paradigms of parameter updating, which inherently faces challenges such as high development costs, poor robustness, training instability, and lack of interpretability. Recently, large language models (LLMs) have demonstrated significant potential across various domains, notably through In-Context Learning (ICL), which enables high-performance task execution with minimal input-label demonstrations and no prior training. This paper aims to replace supervised learning with the emerging ICL paradigm for soft sensor modeling to address existing challenges and explore new avenues for advancement. To achieve this, we propose a novel framework called the Few-shot Uncertainty-aware and self-Explaining Soft Sensor (LLM-FUESS), which includes the Zero-shot Auxiliary Variable Selector (LLM-ZAVS) and the Uncertainty-aware Few-shot Soft Sensor (LLM-UFSS). The LLM-ZAVS retrieves from the Industrial Knowledge Vector Storage to enhance LLMs' domain-specific knowledge, enabling zero-shot auxiliary variable selection. In the LLM-UFSS, we utilize text-based context demonstrations of structured data to prompt LLMs to execute ICL for predicting and propose a context sample retrieval augmentation strategy to improve performance. Additionally, we explored LLMs' AIGC and probabilistic characteristics to propose self-explanation and uncertainty quantification methods for constructing a trustworthy soft sensor. Extensive experiments demonstrate that our method achieved state-of-the-art predictive performance, strong robustness, and flexibility, effectively mitigates training instability found in traditional methods. To the best of our knowledge, this is the first work to establish soft sensor utilizing LLMs.
2501.03300
Method of data forward generation with partial differential equations for machine learning modeling in fluid mechanics
cs.LG physics.flu-dyn
Artificial intelligence (AI) for fluid mechanics has become attractive topic. High-fidelity data is one of most critical issues for the successful applications of AI in fluid mechanics, however, it is expensively obtained or even inaccessible. This study proposes a high-efficient data forward generation method from the partial differential equations (PDEs). Specifically, the solutions of the PDEs are first generated either following a random field (e.g. Gaussian random field, GRF, computational complexity O(NlogN), N is the number of spatial points) or physical laws (e.g. a kind of spectra, computational complexity O(NM), M is the number of modes), then the source terms, boundary conditions and initial conditions are computed to satisfy PDEs. Thus, the data pairs of source terms, boundary conditions and initial conditions with corresponding solutions of PDEs can be constructed. A Poisson neural network (Poisson-NN) embedded in projection method and a wavelet transform convolutional neuro network (WTCNN) embedded in multigrid numerical simulation for solving incompressible Navier-Stokes equations is respectively proposed. The feasibility of generated data for training Poisson-NN and WTCNN is validated. The results indicate that even without any DNS data, the generated data can train these two models with excellent generalization and accuracy. The data following physical laws can significantly improve the convergence rate, generalization and accuracy than that generated following GRF.
2501.03301
Rethinking Byzantine Robustness in Federated Recommendation from Sparse Aggregation Perspective
cs.CR cs.AI cs.DC cs.LG
To preserve user privacy in recommender systems, federated recommendation (FR) based on federated learning (FL) emerges, keeping the personal data on the local client and updating a model collaboratively. Unlike FL, FR has a unique sparse aggregation mechanism, where the embedding of each item is updated by only partial clients, instead of full clients in a dense aggregation of general FL. Recently, as an essential principle of FL, model security has received increasing attention, especially for Byzantine attacks, where malicious clients can send arbitrary updates. The problem of exploring the Byzantine robustness of FR is particularly critical since in the domains applying FR, e.g., e-commerce, malicious clients can be injected easily by registering new accounts. However, existing Byzantine works neglect the unique sparse aggregation of FR, making them unsuitable for our problem. Thus, we make the first effort to investigate Byzantine attacks on FR from the perspective of sparse aggregation, which is non-trivial: it is not clear how to define Byzantine robustness under sparse aggregations and design Byzantine attacks under limited knowledge/capability. In this paper, we reformulate the Byzantine robustness under sparse aggregation by defining the aggregation for a single item as the smallest execution unit. Then we propose a family of effective attack strategies, named Spattack, which exploit the vulnerability in sparse aggregation and are categorized along the adversary's knowledge and capability. Extensive experimental results demonstrate that Spattack can effectively prevent convergence and even break down defenses under a few malicious clients, raising alarms for securing FR systems.
2501.03304
LiLMaps: Learnable Implicit Language Maps
cs.RO cs.LG
One of the current trends in robotics is to employ large language models (LLMs) to provide non-predefined command execution and natural human-robot interaction. It is useful to have an environment map together with its language representation, which can be further utilized by LLMs. Such a comprehensive scene representation enables numerous ways of interaction with the map for autonomously operating robots. In this work, we present an approach that enhances incremental implicit mapping through the integration of vision-language features. Specifically, we (i) propose a decoder optimization technique for implicit language maps which can be used when new objects appear on the scene, and (ii) address the problem of inconsistent vision-language predictions between different viewing positions. Our experiments demonstrate the effectiveness of LiLMaps and solid improvements in performance.
2501.03305
Plant Leaf Disease Detection and Classification Using Deep Learning: A Review and A Proposed System on Bangladesh's Perspective
cs.CV cs.LG
A very crucial part of Bangladeshi people's employment, GDP contribution, and mainly livelihood is agriculture. It plays a vital role in decreasing poverty and ensuring food security. Plant diseases are a serious stumbling block in agricultural production in Bangladesh. At times, humans can't detect the disease from an infected leaf with the naked eye. Using inorganic chemicals or pesticides in plants when it's too late leads in vain most of the time, deposing all the previous labor. The deep-learning technique of leaf-based image classification, which has shown impressive results, can make the work of recognizing and classifying all diseases trouble-less and more precise. In this paper, we've mainly proposed a better model for the detection of leaf diseases. Our proposed paper includes the collection of data on three different kinds of crops: bell peppers, tomatoes, and potatoes. For training and testing the proposed CNN model, the plant leaf disease dataset collected from Kaggle is used, which has 17,430 images. The images are labeled with 14 separate classes of damage. The developed CNN model performs efficiently and could successfully detect and classify the tested diseases. The proposed CNN model may have great potency in crop disease management.
2501.03306
The Robustness of Spiking Neural Networks in Federated Learning with Compression Against Non-omniscient Byzantine Attacks
cs.CR cs.DC cs.LG
Spiking Neural Networks (SNNs), which offer exceptional energy efficiency for inference, and Federated Learning (FL), which offers privacy-preserving distributed training, is a rising area of interest that highly beneficial towards Internet of Things (IoT) devices. Despite this, research that tackles Byzantine attacks and bandwidth limitation in FL-SNNs, both poses significant threats on model convergence and training times, still remains largely unexplored. Going beyond proposing a solution for both of these problems, in this work we highlight the dual benefits of FL-SNNs, against non-omniscient Byzantine adversaries (ones that restrict attackers access to local clients datasets), and greater communication efficiency, over FL-ANNs. Specifically, we discovered that a simple integration of Top-\k{appa} sparsification into the FL apparatus can help leverage the advantages of the SNN models in both greatly reducing bandwidth usage and significantly boosting the robustness of FL training against non-omniscient Byzantine adversaries. Most notably, we saw a massive improvement of roughly 40% accuracy gain in FL-SNNs training under the lethal MinMax attack
2501.03324
Analyzing Bias in Swiss Federal Supreme Court Judgments Using Facebook's Holistic Bias Dataset: Implications for Language Model Training
cs.CL cs.AI
Natural Language Processing (NLP) is vital for computers to process and respond accurately to human language. However, biases in training data can introduce unfairness, especially in predicting legal judgment. This study focuses on analyzing biases within the Swiss Judgment Prediction Dataset (SJP-Dataset). Our aim is to ensure unbiased factual descriptions essential for fair decision making by NLP models in legal contexts. We analyze the dataset using social bias descriptors from the Holistic Bias dataset and employ advanced NLP techniques, including attention visualization, to explore the impact of dispreferred descriptors on model predictions. The study identifies biases and examines their influence on model behavior. Challenges include dataset imbalance and token limits affecting model performance.
2501.03331
Global network control from local information
eess.SY cond-mat.dis-nn cs.SY
In the classical control of network systems, the control actions on a node are determined as a function of the states of all nodes in the network. Motivated by applications where the global state cannot be reconstructed in real time due to limitations in the collection, communication, and processing of data, here we introduce a control approach in which the control actions can be computed as a function of the states of the nodes within a limited state information neighborhood. The trade-off between the control performance and the size of this neighborhood is primarily determined by the condition number of the controllability Gramian. Our theoretical results are supported by simulations on regular and random networks and are further illustrated by an application to the control of power-grid synchronization. We demonstrate that for well-conditioned Gramians, there is no significant loss of control performance as the size of the state information neighborhood is reduced, allowing efficient control of large networks using only local information.
2501.03332
CM3T: Framework for Efficient Multimodal Learning for Inhomogeneous Interaction Datasets
cs.CV
Challenges in cross-learning involve inhomogeneous or even inadequate amount of training data and lack of resources for retraining large pretrained models. Inspired by transfer learning techniques in NLP, adapters and prefix tuning, this paper presents a new model-agnostic plugin architecture for cross-learning, called CM3T, that adapts transformer-based models to new or missing information. We introduce two adapter blocks: multi-head vision adapters for transfer learning and cross-attention adapters for multimodal learning. Training becomes substantially efficient as the backbone and other plugins do not need to be finetuned along with these additions. Comparative and ablation studies on three datasets Epic-Kitchens-100, MPIIGroupInteraction and UDIVA v0.5 show efficacy of this framework on different recording settings and tasks. With only 12.8% trainable parameters compared to the backbone to process video input and only 22.3% trainable parameters for two additional modalities, we achieve comparable and even better results than the state-of-the-art. CM3T has no specific requirements for training or pretraining and is a step towards bridging the gap between a general model and specific practical applications of video classification.
2501.03336
Mobile Augmented Reality Framework with Fusional Localization and Pose Estimation
cs.CV
As a novel way of presenting information, augmented reality (AR) enables people to interact with the physical world in a direct and intuitive way. While there are some mobile AR products implemented with specific hardware at a high cost, the software approaches of AR implementation on mobile platforms(such as smartphones, tablet PC, etc.) are still far from practical use. GPS-based mobile AR systems usually perform poorly due to the inaccurate positioning in the indoor environment. Previous vision-based pose estimation methods need to continuously track predefined markers within a short distance, which greatly degrade user experience. This paper first conducts a comprehensive study of the state-of-the-art AR and localization systems on mobile platforms. Then, we propose an effective indoor mobile AR framework. In the framework, a fusional localization method and a new pose estimation implementation are developed to increase the overall matching rate and thus improving AR display accuracy. Experiments show that our framework has higher performance than approaches purely based on images or Wi-Fi signals. We achieve low average error distances (0.61-0.81m) and accurate matching rates (77%-82%) when the average sampling grid length is set to 0.5m.
2501.03349
FTA-FTL: A Fine-Tuned Aggregation Federated Transfer Learning Scheme for Lithology Microscopic Image Classification
cs.LG cs.AI cs.CV
Lithology discrimination is a crucial activity in characterizing oil reservoirs, and processing lithology microscopic images is an essential technique for investigating fossils and minerals and geological assessment of shale oil exploration. In this way, Deep Learning (DL) technique is a powerful approach for building robust classifier models. However, there is still a considerable challenge to collect and produce a large dataset. Transfer-learning and data augmentation techniques have emerged as popular approaches to tackle this problem. Furthermore, due to different reasons, especially data privacy, individuals, organizations, and industry companies often are not willing to share their sensitive data and information. Federated Learning (FL) has emerged to train a highly accurate central model across multiple decentralized edge servers without transferring sensitive data, preserving sensitive data, and enhancing security. This study involves two phases; the first phase is to conduct Lithology microscopic image classification on a small dataset using transfer learning. In doing so, various pre-trained DL model architectures are comprehensively compared for the classification task. In the second phase, we formulated the classification task to a Federated Transfer Learning (FTL) scheme and proposed a Fine-Tuned Aggregation strategy for Federated Learning (FTA-FTL). In order to perform a comprehensive experimental study, several metrics such as accuracy, f1 score, precision, specificity, sensitivity (recall), and confusion matrix are taken into account. The results are in excellent agreement and confirm the efficiency of the proposed scheme, and show that the proposed FTA-FTL algorithm is capable enough to achieve approximately the same results obtained by the centralized implementation for Lithology microscopic images classification task.
2501.03358
Data integrity vs. inference accuracy in large AIS datasets
cs.CR cs.LG
Automatic Ship Identification Systems (AIS) play a key role in monitoring maritime traffic, providing the data necessary for analysis and decision-making. The integrity of this data is fundamental to the correctness of infer-ence and decision-making in the context of maritime safety, traffic manage-ment and environmental protection. This paper analyzes the impact of data integrity in large AIS datasets, on classification accuracy. It also presents er-ror detection and correction methods and data verification techniques that can improve the reliability of AIS systems. The results show that improving the integrity of AIS data significantly improves the quality of inference, which has a direct impact on operational efficiency and safety at sea.
2501.03360
Quantum Feature-Empowered Deep Classification for Fast Mangrove Mapping
quant-ph cs.CV eess.IV
A mangrove mapping (MM) algorithm is an essential classification tool for environmental monitoring. The recent literature shows that compared with other index-based MM methods that treat pixels as spatially independent, convolutional neural networks (CNNs) are crucial for leveraging spatial continuity information, leading to improved classification performance. In this work, we go a step further to show that quantum features provide radically new information for CNN to further upgrade the classification results. Simply speaking, CNN computes affine-mapping features, while quantum neural network (QNN) offers unitary-computing features, thereby offering a fresh perspective in the final decision-making (classification). To address the challenging MM problem, we design an entangled spatial-spectral quantum feature extraction module. Notably, to ensure that the quantum features contribute genuinely novel information (unaffected by traditional CNN features), we design a separate network track consisting solely of quantum neurons with built-in interpretability. The extracted pure quantum information is then fused with traditional feature information to jointly make the final decision. The proposed quantum-empowered deep network (QEDNet) is very lightweight, so the improvement does come from the cooperation between CNN and QNN (rather than parameter augmentation). Extensive experiments will be conducted to demonstrate the superiority of QEDNet.
2501.03368
Detecting Defective Wafers Via Modular Networks
cs.LG
The growing availability of sensors within semiconductor manufacturing processes makes it feasible to detect defective wafers with data-driven models. Without directly measuring the quality of semiconductor devices, they capture the modalities between diverse sensor readings and can be used to predict key quality indicators (KQI, \textit{e.g.}, roughness, resistance) to detect faulty products, significantly reducing the capital and human cost in maintaining physical metrology steps. Nevertheless, existing models pay little attention to the correlations among different processes for diverse wafer products and commonly struggle with generalizability issues. To enable generic fault detection, in this work, we propose a modular network (MN) trained using time series stage-wise datasets that embodies the structure of the manufacturing process. It decomposes KQI prediction as a combination of stage modules to simulate compositional semiconductor manufacturing, universally enhancing faulty wafer detection among different wafer types and manufacturing processes. Extensive experiments demonstrate the usefulness of our approach, and shed light on how the compositional design provides an interpretable interface for more practical applications.
2501.03370
Advanced Machine Learning Techniques for Social Support Detection on Social Media
cs.CL cs.AI cs.HC cs.LG
The widespread use of social media highlights the need to understand its impact, particularly the role of online social support. This study uses a dataset focused on online social support, which includes binary and multiclass classifications of social support content on social media. The classification of social support is divided into three tasks. The first task focuses on distinguishing between supportive and non-supportive. The second task aims to identify whether the support is directed toward an individual or a group. The third task categorizes the specific type of social support, grouping it into categories such as Nation, LGBTQ, Black people, Women, Religion, and Other (if it does not fit into the previously mentioned categories). To address data imbalances in these tasks, we employed K-means clustering for balancing the dataset and compared the results with the original unbalanced data. Using advanced machine learning techniques, including transformers and zero-shot learning approaches with GPT3, GPT4, and GPT4-o, we predict social support levels in various contexts. The effectiveness of the dataset is evaluated using baseline models across different learning approaches, with transformer-based methods demonstrating superior performance. Additionally, we achieved a 0.4\% increase in the macro F1 score for the second task and a 0.7\% increase for the third task, compared to previous work utilizing traditional machine learning with psycholinguistic and unigram-based TF-IDF values.
2501.03374
License Plate Images Generation with Diffusion Models
cs.CV cs.AI cs.LG
Despite the evident practical importance of license plate recognition (LPR), corresponding research is limited by the volume of publicly available datasets due to privacy regulations such as the General Data Protection Regulation (GDPR). To address this challenge, synthetic data generation has emerged as a promising approach. In this paper, we propose to synthesize realistic license plates (LPs) using diffusion models, inspired by recent advances in image and video generation. In our experiments a diffusion model was successfully trained on a Ukrainian LP dataset, and 1000 synthetic images were generated for detailed analysis. Through manual classification and annotation of the generated images, we performed a thorough study of the model output, such as success rate, character distributions, and type of failures. Our contributions include experimental validation of the efficacy of diffusion models for LP synthesis, along with insights into the characteristics of the generated data. Furthermore, we have prepared a synthetic dataset consisting of 10,000 LP images, publicly available at https://zenodo.org/doi/10.5281/zenodo.13342102. Conducted experiments empirically confirm the usefulness of synthetic data for the LPR task. Despite the initial performance gap between the model trained with real and synthetic data, the expansion of the training data set with pseudolabeled synthetic data leads to an improvement in LPR accuracy by 3% compared to baseline.
2501.03376
Existential Crisis: A Social Robot's Reason for Being
cs.RO cs.AI cs.HC
As Robots become ever more important in our daily lives there's growing need for understanding how they're perceived by people. This study aims to investigate how the user perception of robots is influenced by displays of personality. Using LLMs and speech to text technology, we designed a within-subject study to compare two conditions: a personality-driven robot and a purely task-oriented, personality-neutral robot. Twelve participants, recruited from Socially Intelligent Robotics course at Vrije Universiteit Amsterdam, interacted with a robot Nao tasked with asking them a set of medical questions under both conditions. After completing both interactions, the participants completed a user experience questionnaire measuring their emotional states and robot perception using standardized questionnaires from the SRI and Psychology literature.
2501.03383
The Artificial Scientist -- in-transit Machine Learning of Plasma Simulations
physics.comp-ph cs.DC cs.LG
Increasing HPC cluster sizes and large-scale simulations that produce petabytes of data per run, create massive IO and storage challenges for analysis. Deep learning-based techniques, in particular, make use of these amounts of domain data to extract patterns that help build scientific understanding. Here, we demonstrate a streaming workflow in which simulation data is streamed directly to a machine-learning (ML) framework, circumventing the file system bottleneck. Data is transformed in transit, asynchronously to the simulation and the training of the model. With the presented workflow, data operations can be performed in common and easy-to-use programming languages, freeing the application user from adapting the application output routines. As a proof-of-concept we consider a GPU accelerated particle-in-cell (PIConGPU) simulation of the Kelvin- Helmholtz instability (KHI). We employ experience replay to avoid catastrophic forgetting in learning from this non-steady process in a continual manner. We detail challenges addressed while porting and scaling to Frontier exascale system.
2501.03392
Over-the-Air Fair Federated Learning via Multi-Objective Optimization
cs.LG cs.AI
In federated learning (FL), heterogeneity among the local dataset distributions of clients can result in unsatisfactory performance for some, leading to an unfair model. To address this challenge, we propose an over-the-air fair federated learning algorithm (OTA-FFL), which leverages over-the-air computation to train fair FL models. By formulating FL as a multi-objective minimization problem, we introduce a modified Chebyshev approach to compute adaptive weighting coefficients for gradient aggregation in each communication round. To enable efficient aggregation over the multiple access channel, we derive analytical solutions for the optimal transmit scalars at the clients and the de-noising scalar at the parameter server. Extensive experiments demonstrate the superiority of OTA-FFL in achieving fairness and robust performance compared to existing methods.
2501.03394
Enhanced Importance Sampling through Latent Space Exploration in Normalizing Flows
cs.RO cs.AI cs.LG
Importance sampling is a rare event simulation technique used in Monte Carlo simulations to bias the sampling distribution towards the rare event of interest. By assigning appropriate weights to sampled points, importance sampling allows for more efficient estimation of rare events or tails of distributions. However, importance sampling can fail when the proposal distribution does not effectively cover the target distribution. In this work, we propose a method for more efficient sampling by updating the proposal distribution in the latent space of a normalizing flow. Normalizing flows learn an invertible mapping from a target distribution to a simpler latent distribution. The latent space can be more easily explored during the search for a proposal distribution, and samples from the proposal distribution are recovered in the space of the target distribution via the invertible mapping. We empirically validate our methodology on simulated robotics applications such as autonomous racing and aircraft ground collision avoidance.
2501.03397
DoubleDiffusion: Combining Heat Diffusion with Denoising Diffusion for Generative Learning on 3D Meshes
cs.CV
This paper proposes DoubleDiffusion, a novel framework that combines heat dissipation diffusion and denoising diffusion for direct generative learning on 3D mesh surfaces. Our approach addresses the challenges of generating continuous signal distributions residing on a curve manifold surface. Unlike previous methods that rely on unrolling 3D meshes into 2D or adopting field representations, DoubleDiffusion leverages the Laplacian-Beltrami operator to process features respecting the mesh structure. This combination enables effective geometry-aware signal diffusion across the underlying geometry. As shown in Fig.1, we demonstrate that DoubleDiffusion has the ability to generate RGB signal distributions on complex 3D mesh surfaces and achieves per-category shape-conditioned texture generation across different shape geometry. Our work contributes a new direction in diffusion-based generative modeling on 3D surfaces, with potential applications in the field of 3D asset generation.
2501.03399
Compression of 3D Gaussian Splatting with Optimized Feature Planes and Standard Video Codecs
cs.CV cs.MM
3D Gaussian Splatting is a recognized method for 3D scene representation, known for its high rendering quality and speed. However, its substantial data requirements present challenges for practical applications. In this paper, we introduce an efficient compression technique that significantly reduces storage overhead by using compact representation. We propose a unified architecture that combines point cloud data and feature planes through a progressive tri-plane structure. Our method utilizes 2D feature planes, enabling continuous spatial representation. To further optimize these representations, we incorporate entropy modeling in the frequency domain, specifically designed for standard video codecs. We also propose channel-wise bit allocation to achieve a better trade-off between bitrate consumption and feature plane representation. Consequently, our model effectively leverages spatial correlations within the feature planes to enhance rate-distortion performance using standard, non-differentiable video codecs. Experimental results demonstrate that our method outperforms existing methods in data compactness while maintaining high rendering quality. Our project page is available at https://fraunhoferhhi.github.io/CodecGS
2501.03400
Power System Steady-State Estimation Revisited
math.OC cs.SY eess.SY
In power system steady-state estimation (PSSE), one needs to consider (1) the need for robust statistics, (2) the nonconvex transmission constraints, (3) the fast-varying nature of the inputs, and the corresponding need to track optimal trajectories as closely as possible. In combination, these challenges have not been considered, yet. In this paper, we address all three challenges. The need for robustness (1) is addressed by using an approach based on the so-called Huber model. The non-convexity (2) of the problem, which results in first order methods failing to find global minima, is dealt with by applying global methods. One of these methods is based on a mixed integer quadratic formulation, which provides results of several orders of magnitude better than conventional gradient descent. Lastly, the trajectory tracking (3) is discussed by showing under which conditions the trajectory tracking of the SDP relaxations has meaning.
2501.03402
On the Adversarial Robustness of Benjamini Hochberg
math.ST cs.LG stat.TH
The Benjamini-Hochberg (BH) procedure is widely used to control the false detection rate (FDR) in multiple testing. Applications of this control abound in drug discovery, forensics, anomaly detection, and, in particular, machine learning, ranging from nonparametric outlier detection to out-of-distribution detection and one-class classification methods. Considering this control could be relied upon in critical safety/security contexts, we investigate its adversarial robustness. More precisely, we study under what conditions BH does and does not exhibit adversarial robustness, we present a class of simple and easily implementable adversarial test-perturbation algorithms, and we perform computational experiments. With our algorithms, we demonstrate that there are conditions under which BH's control can be significantly broken with relatively few (even just one) test score perturbation(s), and provide non-asymptotic guarantees on the expected adversarial-adjustment to FDR. Our technical analysis involves a combinatorial reframing of the BH procedure as a ``balls into bins'' process, and drawing a connection to generalized ballot problems to facilitate an information-theoretic approach for deriving non-asymptotic lower bounds.
2501.03403
BoundingDocs: a Unified Dataset for Document Question Answering with Spatial Annotations
cs.CL cs.AI
We present a unified dataset for document Question-Answering (QA), which is obtained combining several public datasets related to Document AI and visually rich document understanding (VRDU). Our main contribution is twofold: on the one hand we reformulate existing Document AI tasks, such as Information Extraction (IE), into a Question-Answering task, making it a suitable resource for training and evaluating Large Language Models; on the other hand, we release the OCR of all the documents and include the exact position of the answer to be found in the document image as a bounding box. Using this dataset, we explore the impact of different prompting techniques (that might include bounding box information) on the performance of open-weight models, identifying the most effective approaches for document comprehension.
2501.03405
A Study of the Efficacy of Generative Flow Networks for Robotics and Machine Fault-Adaptation
cs.RO
Advancements in robotics have opened possibilities to automate tasks in various fields such as manufacturing, emergency response and healthcare. However, a significant challenge that prevents robots from operating in real-world environments effectively is out-of-distribution (OOD) situations, wherein robots encounter unforseen situations. One major OOD situations is when robots encounter faults, making fault adaptation essential for real-world operation for robots. Current state-of-the-art reinforcement learning algorithms show promising results but suffer from sample inefficiency, leading to low adaptation speed due to their limited ability to generalize to OOD situations. Our research is a step towards adding hardware fault tolerance and fast fault adaptability to machines. In this research, our primary focus is to investigate the efficacy of generative flow networks in robotic environments, particularly in the domain of machine fault adaptation. We simulated a robotic environment called Reacher in our experiments. We modify this environment to introduce four distinct fault environments that replicate real-world machines/robot malfunctions. The empirical evaluation of this research indicates that continuous generative flow networks (CFlowNets) indeed have the capability to add adaptive behaviors in machines under adversarial conditions. Furthermore, the comparative analysis of CFlowNets with reinforcement learning algorithms also provides some key insights into the performance in terms of adaptation speed and sample efficiency. Additionally, a separate study investigates the implications of transferring knowledge from pre-fault task to post-fault environments. Our experiments confirm that CFlowNets has the potential to be deployed in a real-world machine and it can demonstrate adaptability in case of malfunctions to maintain functionality.
2501.03406
Low-Order Flow Reconstruction and Uncertainty Quantification in Disturbed Aerodynamics Using Sparse Pressure Measurements
cs.LG physics.flu-dyn
This paper presents a novel machine-learning framework for reconstructing low-order gust-encounter flow field and lift coefficients from sparse, noisy surface pressure measurements. Our study thoroughly investigates the time-varying response of sensors to gust-airfoil interactions, uncovering valuable insights into optimal sensor placement. To address uncertainties in deep learning predictions, we implement probabilistic regression strategies to model both epistemic and aleatoric uncertainties. Epistemic uncertainty, reflecting the model's confidence in its predictions, is modeled using Monte Carlo dropout, as an approximation to the variational inference in the Bayesian framework, treating the neural network as a stochastic entity. On the other hand, aleatoric uncertainty, arising from noisy input measurements, is captured via learned statistical parameters, which propagates measurement noise through the network into the final predictions. Our results showcase the efficacy of this dual uncertainty quantification strategy in accurately predicting aerodynamic behavior under extreme conditions while maintaining computational efficiency, underscoring its potential to improve online sensor-based flow estimation in real-world applications.
2501.03410
ScaleMAI: Accelerating the Development of Trusted Datasets and AI Models
cs.CV
Building trusted datasets is critical for transparent and responsible Medical AI (MAI) research, but creating even small, high-quality datasets can take years of effort from multidisciplinary teams. This process often delays AI benefits, as human-centric data creation and AI-centric model development are treated as separate, sequential steps. To overcome this, we propose ScaleMAI, an agent of AI-integrated data curation and annotation, allowing data quality and AI performance to improve in a self-reinforcing cycle and reducing development time from years to months. We adopt pancreatic tumor detection as an example. First, ScaleMAI progressively creates a dataset of 25,362 CT scans, including per-voxel annotations for benign/malignant tumors and 24 anatomical structures. Second, through progressive human-in-the-loop iterations, ScaleMAI provides Flagship AI Model that can approach the proficiency of expert annotators (30-year experience) in detecting pancreatic tumors. Flagship Model significantly outperforms models developed from smaller, fixed-quality datasets, with substantial gains in tumor detection (+14%), segmentation (+5%), and classification (72%) on three prestigious benchmarks. In summary, ScaleMAI transforms the speed, scale, and reliability of medical dataset creation, paving the way for a variety of impactful, data-driven applications.
2501.03413
SALT: Sales Autocompletion Linked Business Tables Dataset
cs.LG cs.AI cs.DB
Foundation models, particularly those that incorporate Transformer architectures, have demonstrated exceptional performance in domains such as natural language processing and image processing. Adapting these models to structured data, like tables, however, introduces significant challenges. These difficulties are even more pronounced when addressing multi-table data linked via foreign key, which is prevalent in the enterprise realm and crucial for empowering business use cases. Despite its substantial impact, research focusing on such linked business tables within enterprise settings remains a significantly important yet underexplored domain. To address this, we introduce a curated dataset sourced from an Enterprise Resource Planning (ERP) system, featuring extensive linked tables. This dataset is specifically designed to support research endeavors in table representation learning. By providing access to authentic enterprise data, our goal is to potentially enhance the effectiveness and applicability of models for real-world business contexts.
2501.03416
TinySense: A Lighter Weight and More Power-efficient Avionics System for Flying Insect-scale Robots
cs.RO cs.SY eess.SY
In this paper, we investigate the prospects and challenges of sensor suites in achieving autonomous control for flying insect robots (FIRs) weighing less than a gram. FIRs, owing to their minuscule weight and size, offer unparalleled advantages in terms of material cost and scalability. However, their size introduces considerable control challenges, notably high-speed dynamics, restricted power, and limited payload capacity. While there have been notable advancements in developing lightweight sensors, often drawing inspiration from biological systems, no sub-gram aircraft has been able to attain sustained hover without relying on feedback from external sensing such as a motion capture system. The lightest vehicle capable of sustained hover -- the first level of "sensor autonomy" -- is the much larger 28 g Crazyflie. Previous work reported a reduction in size of that vehicle's avionics suite to 187 mg and 21 mW. Here, we report a further reduction in mass and power to only 78.4 mg and 15 mW. We replaced the laser rangefinder with a lighter and more efficient pressure sensor, and built a smaller optic flow sensor around a global-shutter imaging chip. A Kalman Filter (KF) fuses these measurements to estimate the state variables that are needed to control hover: pitch angle, translational velocity, and altitude. Our system achieved performance comparable to that of the Crazyflie's estimator while in flight, with root mean squared errors of 1.573 degrees, 0.186 m/s, and 0.139 m, respectively, relative to motion capture.
2501.03420
Designing Telepresence Robots to Support Place Attachment
cs.HC cs.RO
People feel attached to places that are meaningful to them, which psychological research calls "place attachment." Place attachment is associated with self-identity, self-continuity, and psychological well-being. Even small cues, including videos, images, sounds, and scents, can facilitate feelings of connection and belonging to a place. Telepresence robots that allow people to see, hear, and interact with a remote place have the potential to establish and maintain a connection with places and support place attachment. In this paper, we explore the design space of robotic telepresence to promote place attachment, including how users might be guided in a remote place and whether they experience the environment individually or with others. We prototyped a telepresence robot that allows one or more remote users to visit a place and be guided by a local human guide or a conversational agent. Participants were 38 university alumni who visited their alma mater via the telepresence robot. Our findings uncovered four distinct user personas in the remote experience and highlighted the need for social participation to enhance place attachment. We generated design implications for future telepresence robot design to support people's connections with places of personal significance.
2501.03430
A Self-supervised Diffusion Bridge for MRI Reconstruction
eess.IV cs.CV
Diffusion bridges (DBs) are a class of diffusion models that enable faster sampling by interpolating between two paired image distributions. Training traditional DBs for image reconstruction requires high-quality reference images, which limits their applicability to settings where such references are unavailable. We propose SelfDB as a novel self-supervised method for training DBs directly on available noisy measurements without any high-quality reference images. SelfDB formulates the diffusion process by further sub-sampling the available measurements two additional times and training a neural network to reverse the corresponding degradation process by using the available measurements as the training targets. We validate SelfDB on compressed sensing MRI, showing its superior performance compared to the denoising diffusion models.
2501.03432
Mixture-of-Experts Graph Transformers for Interpretable Particle Collision Detection
cs.LG hep-ph
The Large Hadron Collider at CERN produces immense volumes of complex data from high-energy particle collisions, demanding sophisticated analytical techniques for effective interpretation. Neural Networks, including Graph Neural Networks, have shown promise in tasks such as event classification and object identification by representing collisions as graphs. However, while Graph Neural Networks excel in predictive accuracy, their "black box" nature often limits their interpretability, making it difficult to trust their decision-making processes. In this paper, we propose a novel approach that combines a Graph Transformer model with Mixture-of-Expert layers to achieve high predictive performance while embedding interpretability into the architecture. By leveraging attention maps and expert specialization, the model offers insights into its internal decision-making, linking predictions to physics-informed features. We evaluate the model on simulated events from the ATLAS experiment, focusing on distinguishing rare Supersymmetric signal events from Standard Model background. Our results highlight that the model achieves competitive classification accuracy while providing interpretable outputs that align with known physics, demonstrating its potential as a robust and transparent tool for high-energy physics data analysis. This approach underscores the importance of explainability in machine learning methods applied to high energy physics, offering a path toward greater trust in AI-driven discoveries.
2501.03437
DAMAGE: Detecting Adversarially Modified AI Generated Text
cs.CL
AI humanizers are a new class of online software tools meant to paraphrase and rewrite AI-generated text in a way that allows them to evade AI detection software. We study 19 AI humanizer and paraphrasing tools and qualitatively assess their effects and faithfulness in preserving the meaning of the original text. We show that many existing AI detectors fail to detect humanized text. Finally, we demonstrate a robust model that can detect humanized AI text while maintaining a low false positive rate using a data-centric augmentation approach. We attack our own detector, training our own fine-tuned model optimized against our detector's predictions, and show that our detector's cross-humanizer generalization is sufficient to remain robust to this attack.
2501.03441
Finding A Voice: Evaluating African American Dialect Generation for Chatbot Technology
cs.CL
As chatbots become increasingly integrated into everyday tasks, designing systems that accommodate diverse user populations is crucial for fostering trust, engagement, and inclusivity. This study investigates the ability of contemporary Large Language Models (LLMs) to generate African American Vernacular English (AAVE) and evaluates the impact of AAVE usage on user experiences in chatbot applications. We analyze the performance of three LLM families (Llama, GPT, and Claude) in producing AAVE-like utterances at varying dialect intensities and assess user preferences across multiple domains, including healthcare and education. Despite LLMs' proficiency in generating AAVE-like language, findings indicate that AAVE-speaking users prefer Standard American English (SAE) chatbots, with higher levels of AAVE correlating with lower ratings for a variety of characteristics, including chatbot trustworthiness and role appropriateness. These results highlight the complexities of creating inclusive AI systems and underscore the need for further exploration of diversity to enhance human-computer interactions.
2501.03443
Optimization Learning
math.OC cs.AI
This article introduces the concept of optimization learning, a methodology to design optimization proxies that learn the input/output mapping of parametric optimization problems. These optimization proxies are trustworthy by design: they compute feasible solutions to the underlying optimization problems, provide quality guarantees on the returned solutions, and scale to large instances. Optimization proxies are differentiable programs that combine traditional deep learning technology with repair or completion layers to produce feasible solutions. The article shows that optimization proxies can be trained end-to-end in a self-supervised way. It presents methodologies to provide performance guarantees and to scale optimization proxies to large-scale optimization problems. The potential of optimization proxies is highlighted through applications in power systems and, in particular, real-time risk assessment and security-constrained optimal power flow.
2501.03445
Physics-Constrained Generative Artificial Intelligence for Rapid Takeoff Trajectory Design
cs.LG
To aid urban air mobility (UAM), electric vertical takeoff and landing (eVTOL) aircraft are being targeted. Conventional multidisciplinary analysis and optimization (MDAO) can be expensive, while surrogate-based optimization can struggle with challenging physical constraints. This work proposes physics-constrained generative adversarial networks (physicsGAN), to intelligently parameterize the takeoff control profiles of an eVTOL aircraft and to transform the original design space to a feasible space. Specifically, the transformed feasible space refers to a space where all designs directly satisfy all design constraints. The physicsGAN-enabled surrogate-based takeoff trajectory design framework was demonstrated on the Airbus A3 Vahana. The physicsGAN generated only feasible control profiles of power and wing angle in the feasible space with around 98.9% of designs satisfying all constraints. The proposed design framework obtained 99.6% accuracy compared with simulation-based optimal design and took only 2.2 seconds, which reduced the computational time by around 200 times. Meanwhile, data-driven GAN-enabled surrogate-based optimization took 21.9 seconds using a derivative-free optimizer, which was around an order of magnitude slower than the proposed framework. Moreover, the data-driven GAN-based optimization using gradient-based optimizers could not consistently find the optimal design during random trials and got stuck in an infeasible region, which is problematic in real practice. Therefore, the proposed physicsGAN-based design framework outperformed data-driven GAN-based design to the extent of efficiency (2.2 seconds), optimality (99.6% accurate), and feasibility (100% feasible). According to the literature review, this is the first physics-constrained generative artificial intelligence enabled by surrogate models.
2501.03448
Optimizing Value of Learning in Task-Oriented Federated Meta-Learning Systems
cs.LG
Federated Learning (FL) has gained significant attention in recent years due to its distributed nature and privacy preserving benefits. However, a key limitation of conventional FL is that it learns and distributes a common global model to all participants, which fails to provide customized solutions for diverse task requirements. Federated meta-learning (FML) offers a promising solution to this issue by enabling devices to finetune local models after receiving a shared meta-model from the server. In this paper, we propose a task-oriented FML framework over non-orthogonal multiple access (NOMA) networks. A novel metric, termed value of learning (VoL), is introduced to assess the individual training needs across devices. Moreover, a task-level weight (TLW) metric is defined based on task requirements and fairness considerations, guiding the prioritization of edge devices during FML training. The formulated problem, to maximize the sum of TLW-based VoL across devices, forms a non-convex mixed-integer non-linear programming (MINLP) challenge, addressed here using a parameterized deep Q-network (PDQN) algorithm to handle both discrete and continuous variables. Simulation results demonstrate that our approach significantly outperforms baseline schemes, underscoring the advantages of the proposed framework.
2501.03449
Feasibility of short blocklength Reed-Muller codes for physical layer security in real environment
cs.IT cs.CR eess.SP math.IT
In this paper, we investigate the application of Reed-Muller (RM) codes for Physical-layer security in a real world wiretap channel scenario. Utilizing software-defined radios (SDRs) in a real indoor environment, we implement a coset coding scheme that leverages the hierarchical structure of RM codes to secure data transmission. The generator matrix of the RM code is used to partition codewords into cosets in the usual way, where each message corresponds to a unique coset, and auxiliary bits select specific codewords within each coset. This approach enables the legitimate receiver (Bob) can decode the transmitted message with minimal information leakage to eavesdropper (Eve) thus protecting the confidentiality of the communication with the help of coset structure. Mutual information neural estimation (MINE) is used to quantify information leakage and validate the effectiveness of the scheme. Experimental results indicate that RM codes can achieve robust security even in practical environments affected by real-world channel impairments. These findings demonstrate the potential of RM codes as an efficient solution for physical-layer security, particularly for applications that require low latency and short blocklengths.
2501.03451
Structure-Preference Enabled Graph Embedding Generation under Differential Privacy
stat.ML cs.LG cs.SI
Graph embedding generation techniques aim to learn low-dimensional vectors for each node in a graph and have recently gained increasing research attention. Publishing low-dimensional node vectors enables various graph analysis tasks, such as structural equivalence and link prediction. Yet, improper publication opens a backdoor to malicious attackers, who can infer sensitive information of individuals from the low-dimensional node vectors. Existing methods tackle this issue by developing deep graph learning models with differential privacy (DP). However, they often suffer from large noise injections and cannot provide structural preferences consistent with mining objectives. Recently, skip-gram based graph embedding generation techniques are widely used due to their ability to extract customizable structures. Based on skip-gram, we present SE-PrivGEmb, a structure-preference enabled graph embedding generation under DP. For arbitrary structure preferences, we design a unified noise tolerance mechanism via perturbing non-zero vectors. This mechanism mitigates utility degradation caused by high sensitivity. By carefully designing negative sampling probabilities in skip-gram, we theoretically demonstrate that skip-gram can preserve arbitrary proximities, which quantify structural features in graphs. Extensive experiments show that our method outperforms existing state-of-the-art methods under structural equivalence and link prediction tasks.
2501.03456
Text to Band Gap: Pre-trained Language Models as Encoders for Semiconductor Band Gap Prediction
cs.CL cond-mat.mtrl-sci
In this study, we explore the use of a transformer-based language model as an encoder to predict the band gaps of semiconductor materials directly from their text descriptions. Quantum chemistry simulations, including Density Functional Theory (DFT), are computationally intensive and time-consuming, which limits their practicality for high-throughput material screening, particularly for complex systems. Shallow machine learning (ML) models, while effective, often require extensive data preprocessing to convert non-numerical material properties into numerical inputs. In contrast, our approach leverages textual data directly, bypassing the need for complex feature engineering. We generate material descriptions in two formats: formatted strings combining features and natural language text generated using the ChatGPT API. We demonstrate that the RoBERTa model, pre-trained on natural language processing tasks, performs effectively as an encoder for prediction tasks. With minimal fine-tuning, it achieves a mean absolute error (MAE) of approximately 0.33 eV, performing better than shallow machine learning models such as Support Vector Regression, Random Forest, and XGBoost. Even when only the linear regression head is trained while keeping the RoBERTa encoder layers frozen, the accuracy remains nearly identical to that of the fully trained model. This demonstrates that the pre-trained RoBERTa encoder is highly adaptable for processing domain-specific text related to material properties, such as the band gap, significantly reducing the need for extensive retraining. This study highlights the potential of transformer-based language models to serve as efficient and versatile encoders for semiconductor materials property prediction tasks.
2501.03458
Activating Associative Disease-Aware Vision Token Memory for LLM-Based X-ray Report Generation
eess.IV cs.AI cs.CV
X-ray image based medical report generation achieves significant progress in recent years with the help of the large language model, however, these models have not fully exploited the effective information in visual image regions, resulting in reports that are linguistically sound but insufficient in describing key diseases. In this paper, we propose a novel associative memory-enhanced X-ray report generation model that effectively mimics the process of professional doctors writing medical reports. It considers both the mining of global and local visual information and associates historical report information to better complete the writing of the current report. Specifically, given an X-ray image, we first utilize a classification model along with its activation maps to accomplish the mining of visual regions highly associated with diseases and the learning of disease query tokens. Then, we employ a visual Hopfield network to establish memory associations for disease-related tokens, and a report Hopfield network to retrieve report memory information. This process facilitates the generation of high-quality reports based on a large language model and achieves state-of-the-art performance on multiple benchmark datasets, including the IU X-ray, MIMIC-CXR, and Chexpert Plus. The source code of this work is released on \url{https://github.com/Event-AHU/Medical_Image_Analysis}.
2501.03461
Radar Signal Recognition through Self-Supervised Learning and Domain Adaptation
cs.LG cs.AI eess.SP
Automatic radar signal recognition (RSR) plays a pivotal role in electronic warfare (EW), as accurately classifying radar signals is critical for informing decision-making processes. Recent advances in deep learning have shown significant potential in improving RSR performance in domains with ample annotated data. However, these methods fall short in EW scenarios where annotated RF data are scarce or impractical to obtain. To address these challenges, we introduce a self-supervised learning (SSL) method which utilises masked signal modelling and RF domain adaption to enhance RSR performance in environments with limited RF samples and labels. Specifically, we investigate pre-training masked autoencoders (MAE) on baseband in-phase and quadrature (I/Q) signals from various RF domains and subsequently transfer the learned representation to the radar domain, where annotated data are limited. Empirical results show that our lightweight self-supervised ResNet model with domain adaptation achieves up to a 17.5% improvement in 1-shot classification accuracy when pre-trained on in-domain signals (i.e., radar signals) and up to a 16.31% improvement when pre-trained on out-of-domain signals (i.e., comm signals), compared to its baseline without SSL. We also provide reference results for several MAE designs and pre-training strategies, establishing a new benchmark for few-shot radar signal classification.
2501.03462
ISSR: Iterative Selection with Self-Review for Vocabulary Test Distractor Generation
cs.CL
Vocabulary acquisition is essential to second language learning, as it underpins all core language skills. Accurate vocabulary assessment is particularly important in standardized exams, where test items evaluate learners' comprehension and contextual use of words. Previous research has explored methods for generating distractors to aid in the design of English vocabulary tests. However, current approaches often rely on lexical databases or predefined rules, and frequently produce distractors that risk invalidating the question by introducing multiple correct options. In this study, we focus on English vocabulary questions from Taiwan's university entrance exams. We analyze student response distributions to gain insights into the characteristics of these test items and provide a reference for future research. Additionally, we identify key limitations in how large language models (LLMs) support teachers in generating distractors for vocabulary test design. To address these challenges, we propose the iterative selection with self-review (ISSR) framework, which makes use of a novel LLM-based self-review mechanism to ensure that the distractors remain valid while offering diverse options. Experimental results show that ISSR achieves promising performance in generating plausible distractors, and the self-review mechanism effectively filters out distractors that could invalidate the question.
2501.03464
LHGNN: Local-Higher Order Graph Neural Networks For Audio Classification and Tagging
cs.SD cs.AI eess.AS
Transformers have set new benchmarks in audio processing tasks, leveraging self-attention mechanisms to capture complex patterns and dependencies within audio data. However, their focus on pairwise interactions limits their ability to process the higher-order relations essential for identifying distinct audio objects. To address this limitation, this work introduces the Local- Higher Order Graph Neural Network (LHGNN), a graph based model that enhances feature understanding by integrating local neighbourhood information with higher-order data from Fuzzy C-Means clusters, thereby capturing a broader spectrum of audio relationships. Evaluation of the model on three publicly available audio datasets shows that it outperforms Transformer-based models across all benchmarks while operating with substantially fewer parameters. Moreover, LHGNN demonstrates a distinct advantage in scenarios lacking ImageNet pretraining, establishing its effectiveness and efficiency in environments where extensive pretraining data is unavailable.
2501.03465
Extending Internet Access Over LoRa for Internet of Things and Critical Applications
cs.NI cs.CY cs.SY eess.SY
LoRa bridges the gap between remote locations and mainstream networks, enabling large-scale Internet of Things (IoT) deployments. Despite the recent advancements around LoRa, Internet access over this technology is still largely unexplored. Most existing solutions only handle packets within the local LoRa network and do not interact with web applications. This limits the scalability and the ability to deliver essential web services in disconnected regions. This work proposes and implements ILoRa to extend the public Internet to disconnected areas for essential service delivery. ILoRa enables accessing Application Programming Interfaces (APIs) and web pages on the Internet over a LoRa backbone network. It comprises a ILoRa coordinator code (ICN) and access point nodes (APNs). The ICN interfaces the LoRa network with the public Internet and interprets content. The APN tethers a WiFi hotspot to which devices connect and access the web content. This work further proposes data handling methods for ICNs and APNs. An actual hardware-based implementation validates the proposed system. The implementation achieves a throughput of 1.06 kbps tested for an Internet-based API returning JSON data of 930 B. Furthermore, the APN consumed approximately $0.162$A current, and the resource utilization on the ICN was minimal.
2501.03466
DGSSA: Domain generalization with structural and stylistic augmentation for retinal vessel segmentation
eess.IV cs.CV
Retinal vascular morphology is crucial for diagnosing diseases such as diabetes, glaucoma, and hypertension, making accurate segmentation of retinal vessels essential for early intervention. Traditional segmentation methods assume that training and testing data share similar distributions, which can lead to poor performance on unseen domains due to domain shifts caused by variations in imaging devices and patient demographics. This paper presents a novel approach, DGSSA, for retinal vessel image segmentation that enhances model generalization by combining structural and style augmentation strategies. We utilize a space colonization algorithm to generate diverse vascular-like structures that closely mimic actual retinal vessels, which are then used to generate pseudo-retinal images with an improved Pix2Pix model, allowing the segmentation model to learn a broader range of structure distributions. Additionally, we utilize PixMix to implement random photometric augmentations and introduce uncertainty perturbations, thereby enriching stylistic diversity and significantly enhancing the model's adaptability to varying imaging conditions. Our framework has been rigorously evaluated on four challenging datasets-DRIVE, CHASEDB, HRF, and STARE-demonstrating state-of-the-art performance that surpasses existing methods. This validates the effectiveness of our proposed approach, highlighting its potential for clinical application in automated retinal vessel analysis.
2501.03467
FRESHR-GSI: A Generalized Safety Model and Evaluation Framework for Mobile Robots in Multi-Human Environments
cs.RO cs.HC
Human safety is critical in applications involving close human-robot interactions (HRI) and is a key aspect of physical compatibility between humans and robots. While measures of human safety in HRI exist, these mainly target industrial settings involving robotic manipulators. Less attention has been paid to settings where mobile robots and humans share the space. This paper introduces a new robot-centered directional framework of human safety. It is particularly useful for evaluating mobile robots as they operate in environments populated by multiple humans. The framework integrates several key metrics, such as each human's relative distance, speed, and orientation. The core novelty lies in the framework's flexibility to accommodate different application requirements while allowing for both the robot-centered and external observer points of view. We instantiate the framework by using RGB-D based vision integrated with a deep learning-based human detection pipeline to yield a generalized safety index (GSI) that instantaneously assesses human safety. We evaluate GSI's capability of producing appropriate, robust, and fine-grained safety measures in real-world experimental scenarios and compare its performance with extant safety models.
2501.03468
MTRAG: A Multi-Turn Conversational Benchmark for Evaluating Retrieval-Augmented Generation Systems
cs.CL cs.AI
Retrieval-augmented generation (RAG) has recently become a very popular task for Large Language Models (LLMs). Evaluating them on multi-turn RAG conversations, where the system is asked to generate a response to a question in the context of a preceding conversation is an important and often overlooked task with several additional challenges. We present MTRAG: an end-to-end human-generated multi-turn RAG benchmark that reflects several real-world properties across diverse dimensions for evaluating the full RAG pipeline. MTRAG contains 110 conversations averaging 7.7 turns each across four domains for a total of 842 tasks. We also explore automation paths via synthetic data and LLM-as-a-Judge evaluation. Our human and automatic evaluations show that even state-of-the-art LLM RAG systems struggle on MTRAG. We demonstrate the need for strong retrieval and generation systems that can handle later turns, unanswerable questions, non-standalone questions, and multiple domains. MTRAG is available at https://github.com/ibm/mt-rag-benchmark.
2501.03469
Information-Maximized Soft Variable Discretization for Self-Supervised Image Representation Learning
cs.CV
Self-supervised learning (SSL) has emerged as a crucial technique in image processing, encoding, and understanding, especially for developing today's vision foundation models that utilize large-scale datasets without annotations to enhance various downstream tasks. This study introduces a novel SSL approach, Information-Maximized Soft Variable Discretization (IMSVD), for image representation learning. Specifically, IMSVD softly discretizes each variable in the latent space, enabling the estimation of their probability distributions over training batches and allowing the learning process to be directly guided by information measures. Motivated by the MultiView assumption, we propose an information-theoretic objective function to learn transform-invariant, non-travail, and redundancy-minimized representation features. We then derive a joint-cross entropy loss function for self-supervised image representation learning, which theoretically enjoys superiority over the existing methods in reducing feature redundancy. Notably, our non-contrastive IMSVD method statistically performs contrastive learning. Extensive experimental results demonstrate the effectiveness of IMSVD on various downstream tasks in terms of both accuracy and efficiency. Thanks to our variable discretization, the embedding features optimized by IMSVD offer unique explainability at the variable level. IMSVD has the potential to be adapted to other learning paradigms. Our code is publicly available at https://github.com/niuchuangnn/IMSVD.
2501.03471
Hyperbolic Binary Neural Network
cs.LG cs.CV
Binary Neural Network (BNN) converts full-precision weights and activations into their extreme 1-bit counterparts, making it particularly suitable for deployment on lightweight mobile devices. While binary neural networks are typically formulated as a constrained optimization problem and optimized in the binarized space, general neural networks are formulated as an unconstrained optimization problem and optimized in the continuous space. This paper introduces the Hyperbolic Binary Neural Network (HBNN) by leveraging the framework of hyperbolic geometry to optimize the constrained problem. Specifically, we transform the constrained problem in hyperbolic space into an unconstrained one in Euclidean space using the Riemannian exponential map. On the other hand, we also propose the Exponential Parametrization Cluster (EPC) method, which, compared to the Riemannian exponential map, shrinks the segment domain based on a diffeomorphism. This approach increases the probability of weight flips, thereby maximizing the information gain in BNNs. Experimental results on CIFAR10, CIFAR100, and ImageNet classification datasets with VGGsmall, ResNet18, and ResNet34 models illustrate the superior performance of our HBNN over state-of-the-art methods.
2501.03475
Reading with Intent -- Neutralizing Intent
cs.CL cs.AI cs.LG
Queries to large language models (LLMs) can be divided into two parts: the instruction/question and the accompanying context. The context for retrieval-augmented generation (RAG) systems in most benchmarks comes from Wikipedia or Wikipedia-like texts which are written in a neutral and factual tone. However, when RAG systems retrieve internet-based content, they encounter text with diverse tones and linguistic styles, introducing challenges for downstream tasks. The Reading with Intent task addresses this issue by evaluating how varying tones in context passages affect model performance. Building on prior work that focused on sarcasm, we extend this paradigm by constructing a dataset where context passages are transformed to $11$ distinct emotions using a better synthetic data generation approach. Using this dataset, we train an emotion translation model to systematically adapt passages to specified emotional tones. The human evaluation shows that the LLM fine-tuned to become the emotion-translator benefited from the synthetically generated data. Finally, the emotion-translator is used in the Reading with Intent task to transform the passages to a neutral tone. By neutralizing the passages, it mitigates the challenges posed by sarcastic passages and improves overall results on this task by about $3\%$.
2501.03477
A study on performance limitations in Federated Learning
cs.LG
Increasing privacy concerns and unrestricted access to data lead to the development of a novel machine learning paradigm called Federated Learning (FL). FL borrows many of the ideas from distributed machine learning, however, the challenges associated with federated learning makes it an interesting engineering problem since the models are trained on edge devices. It was introduced in 2016 by Google, and since then active research is being carried out in different areas within FL such as federated optimization algorithms, model and update compression, differential privacy, robustness, and attacks, federated GANs and privacy preserved personalization. There are many open challenges in the development of such federated machine learning systems and this project will be focusing on the communication bottleneck and data Non IID-ness, and its effect on the performance of the models. These issues are characterized on a baseline model, model performance is evaluated, and discussions are made to overcome these issues.
2501.03479
Women, Infamous, and Exotic Beings: What Honorific Usages in Wikipedia Reveal about the Socio-Cultural Norms
cs.CL
Honorifics serve as powerful linguistic markers that reflect social hierarchies and cultural values. This paper presents a large-scale, cross-linguistic exploration of usage of honorific pronouns in Bengali and Hindi Wikipedia articles, shedding light on how socio-cultural factors shape language. Using LLM (GPT-4o), we annotated 10, 000 articles of real and fictional beings in each language for several sociodemographic features such as gender, age, fame, and exoticness, and the use of honorifics. We find that across all feature combinations, use of honorifics is consistently more common in Bengali than Hindi. For both languages, the use non-honorific pronouns is more commonly observed for infamous, juvenile, and exotic beings. Notably, we observe a gender bias in use of honorifics in Hindi, with men being more commonly referred to with honorifics than women.
2501.03482
VOILA: Complexity-Aware Universal Segmentation of CT images by Voxel Interacting with Language
cs.CV
Satisfactory progress has been achieved recently in universal segmentation of CT images. Following the success of vision-language methods, there is a growing trend towards utilizing text prompts and contrastive learning to develop universal segmentation models. However, there exists a significant imbalance in information density between 3D images and text prompts. Moreover, the standard fully connected layer segmentation approach faces significant challenges in handling multiple classes and exhibits poor generalizability. To address these challenges, we propose the VOxel Interacting with LAnguage method (VOILA) for universal CT image segmentation. Initially, we align voxels and language into a shared representation space and classify voxels on the basis of cosine similarity. Subsequently, we develop the Voxel-Language Interaction framework to mitigate the impact of class imbalance caused by foreground-background discrepancies and variations in target volumes. Furthermore, a Complexity-Aware Sampling method is proposed to focus on region hard to segment, achieved by generating pseudo-heatmaps from a trainable Gaussian mixture distribution. Our results indicate the proposed VOILA is capable to achieve improved performance with reduced parameters and computational cost during training. Furthermore, it demonstrates significant generalizability across diverse datasets without additional fine-tuning.
2501.03486
Align-Pro: A Principled Approach to Prompt Optimization for LLM Alignment
cs.LG cs.AI
The alignment of large language models (LLMs) with human values is critical as these models become increasingly integrated into various societal and decision-making processes. Traditional methods, such as reinforcement learning from human feedback (RLHF), achieve alignment by fine-tuning model parameters, but these approaches are often computationally expensive and impractical when models are frozen or inaccessible for parameter modification. In contrast, prompt optimization is a viable alternative to RLHF for LLM alignment. While the existing literature has shown empirical promise of prompt optimization, its theoretical underpinning remains under-explored. We address this gap by formulating prompt optimization as an optimization problem and try to provide theoretical insights into the optimality of such a framework. To analyze the performance of the prompt optimization, we study theoretical suboptimality bounds and provide insights in terms of how prompt optimization depends upon the given prompter and target model. We also provide empirical validation through experiments on various datasets, demonstrating that prompt optimization can effectively align LLMs, even when parameter fine-tuning is not feasible.
2501.03489
Entropy-Guided Attention for Private LLMs
cs.LG cs.CR
The pervasiveness of proprietary language models has raised critical privacy concerns, necessitating advancements in private inference (PI), where computations are performed directly on encrypted data without revealing users' sensitive information. While PI offers a promising solution, its practical deployment is hindered by substantial communication and latency overheads, primarily stemming from nonlinear operations. To address this, we introduce an information-theoretic framework to characterize the role of nonlinearities in decoder-only language models, laying a principled foundation for optimizing transformer-architectures tailored to the demands of PI. By leveraging Shannon's entropy as a quantitative measure, we uncover the previously unexplored dual significance of nonlinearities: beyond ensuring training stability, they are crucial for maintaining attention head diversity. Specifically, we find that their removal triggers two critical failure modes: {\em entropy collapse} in deeper layers that destabilizes training, and {\em entropic overload} in earlier layers that leads to under-utilization of Multi-Head Attention's (MHA) representational capacity. We propose an entropy-guided attention mechanism paired with a novel entropy regularization technique to mitigate entropic overload. Additionally, we explore PI-friendly alternatives to layer normalization for preventing entropy collapse and stabilizing the training of LLMs with reduced-nonlinearities. Our study bridges the gap between information theory and architectural design, establishing entropy dynamics as a principled guide for developing efficient PI architectures. The code and implementation are available at https://github.com/Nandan91/entropy-guided-attention-llm
2501.03490
SceneBooth: Diffusion-based Framework for Subject-preserved Text-to-Image Generation
cs.CV
Due to the demand for personalizing image generation, subject-driven text-to-image generation method, which creates novel renditions of an input subject based on text prompts, has received growing research interest. Existing methods often learn subject representation and incorporate it into the prompt embedding to guide image generation, but they struggle with preserving subject fidelity. To solve this issue, this paper approaches a novel framework named SceneBooth for subject-preserved text-to-image generation, which consumes inputs of a subject image, object phrases and text prompts. Instead of learning the subject representation and generating a subject, our SceneBooth fixes the given subject image and generates its background image guided by the text prompts. To this end, our SceneBooth introduces two key components, i.e., a multimodal layout generation module and a background painting module. The former determines the position and scale of the subject by generating appropriate scene layouts that align with text captions, object phrases, and subject visual information. The latter integrates two adapters (ControlNet and Gated Self-Attention) into the latent diffusion model to generate a background that harmonizes with the subject guided by scene layouts and text descriptions. In this manner, our SceneBooth ensures accurate preservation of the subject's appearance in the output. Quantitative and qualitative experimental results demonstrate that SceneBooth significantly outperforms baseline methods in terms of subject preservation, image harmonization and overall quality.
2501.03491
Can LLMs Design Good Questions Based on Context?
cs.CL cs.AI
This paper evaluates questions generated by LLMs from context, comparing them to human-generated questions across six dimensions. We introduce an automated LLM-based evaluation method, focusing on aspects like question length, type, context coverage, and answerability. Our findings highlight unique characteristics of LLM-generated questions, contributing insights that can support further research in question quality and downstream applications.
2501.03492
Multi-Source Urban Traffic Flow Forecasting with Drone and Loop Detector Data
cs.LG
Traffic forecasting is a fundamental task in transportation research, however the scope of current research has mainly focused on a single data modality of loop detectors. Recently, the advances in Artificial Intelligence and drone technologies have made possible novel solutions for efficient, accurate and flexible aerial observations of urban traffic. As a promising traffic monitoring approach, drone-captured data can create an accurate multi-sensor mobility observatory for large-scale urban networks, when combined with existing infrastructure. Therefore, this paper investigates the problem of multi-source traffic speed prediction, simultaneously using drone and loop detector data. A simple yet effective graph-based model HiMSNet is proposed to integrate multiple data modalities and learn spatio-temporal correlations. Detailed analysis shows that predicting accurate segment-level speed is more challenging than the regional speed, especially under high-demand scenarios with heavier congestions and varying traffic dynamics. Utilizing both drone and loop detector data, the prediction accuracy can be improved compared to single-modality cases, when the sensors have lower coverages and are subject to noise. Our simulation study based on vehicle trajectories in a real urban road network has highlighted the added value of integrating drones in traffic forecasting and monitoring.
2501.03495
Textualize Visual Prompt for Image Editing via Diffusion Bridge
cs.CV cs.LG
Visual prompt, a pair of before-and-after edited images, can convey indescribable imagery transformations and prosper in image editing. However, current visual prompt methods rely on a pretrained text-guided image-to-image generative model that requires a triplet of text, before, and after images for retraining over a text-to-image model. Such crafting triplets and retraining processes limit the scalability and generalization of editing. In this paper, we present a framework based on any single text-to-image model without reliance on the explicit image-to-image model thus enhancing the generalizability and scalability. Specifically, by leveraging the probability-flow ordinary equation, we construct a diffusion bridge to transfer the distribution between before-and-after images under the text guidance. By optimizing the text via the bridge, the framework adaptively textualizes the editing transformation conveyed by visual prompts into text embeddings without other models. Meanwhile, we introduce differential attention control during text optimization, which disentangles the text embedding from the invariance of the before-and-after images and makes it solely capture the delicate transformation and generalize to edit various images. Experiments on real images validate competitive results on the generalization, contextual coherence, and high fidelity for delicate editing with just one image pair as the visual prompt.
2501.03496
A Unified Attack Detection Strategy for Multi-Agent Systems over Transient and Steady Stages
eess.SY cs.SY
This paper proposes a unified detection strategy against three kinds of attacks for multi-agent systems (MASs) which is applicable to both transient and steady stages. For attacks on the communication layer, a watermarking-based detection scheme with KullbackLeibler (KL) divergence is designed. Different from traditional communication schemes, each agent transmits a message set containing two state values with different types of watermarking. It is found that the detection performance is determined by the relevant parameters of the watermarking signal. Unlike the existing detection manoeuvres, such a scheme is capable of transient and steady stages. For attacks on the agent layer, a convergence rate related detection approach is put forward. It is shown that the resilience of the considered system is characterized by the coefficient and offset of the envelope. For hybrid attacks, based on the above detection mechanisms, a general framework resorting to trusted agents is presented, which requires weaker graph conditions and less information transmission. Finally, an example associated with the platooning of connected vehicles is given to support the theoretical results.
2501.03499
Can Deep Learning Trigger Alerts from Mobile-Captured Images?
cs.CV cs.AI
Our research presents a comprehensive approach to leveraging mobile camera image data for real-time air quality assessment and recommendation. We develop a regression-based Convolutional Neural Network model and tailor it explicitly for air quality prediction by exploiting the inherent relationship between output parameters. As a result, the Mean Squared Error of 0.0077 and 0.0112 obtained for 2 and 5 pollutants respectively outperforms existing models. Furthermore, we aim to verify the common practice of augmenting the original dataset with a view to introducing more variation in the training phase. It is one of our most significant contributions that our experimental results demonstrate minimal accuracy differences between the original and augmented datasets. Finally, a real-time, user-friendly dashboard is implemented which dynamically displays the Air Quality Index and pollutant values derived from captured mobile camera images. Users' health conditions are considered to recommend whether a location is suitable based on current air quality metrics. Overall, this research contributes to verification of data augmentation techniques, CNN-based regression modelling for air quality prediction, and user-centric air quality monitoring through mobile technology. The proposed system offers practical solutions for individuals to make informed environmental health and well-being decisions.
2501.03503
Resilient Distributed Control for Uncertain Nonlinear Interconnected Systems under Network Anomaly
eess.SY cs.SY
We address a distributed adaptive control methodology for nonlinear interconnected systems possibly affected by network anomalies. In the framework of adaptive approximation, the distributed controller and parameter estimator are designed by exploiting a backstepping approach. The stability of the distributed control system under anomalies is analyzed, where both local and neighboring anomaly effects are considered. To quantify the resilience of the interconnected system under the action of network anomalies, we derive bounds on the duration of each anomaly and the resting time between two consecutive anomalies. Specifically, when each anomaly duration is smaller than our designed upper bound, the interconnected system controlled by the distributed approximation-based controller remains asymptotically stable. Moreover, if the resting time between two consecutive anomalies is larger than the proposed bound, then all signals of the control system are guaranteed to be bounded. In the paper, we show that under the action of the proposed distributed adaptive controller, the interconnected system remains stable in the presence of network anomalies, with both the qualitative and quantitative resilient conditions. Extensive simulation results show the effectiveness of our theoretical results.
2501.03507
An Empirical Study of Accuracy-Robustness Tradeoff and Training Efficiency in Self-Supervised Learning
cs.CV cs.LG
Self-supervised learning (SSL) has significantly advanced image representation learning, yet efficiency challenges persist, particularly with adversarial training. Many SSL methods require extensive epochs to achieve convergence, a demand further amplified in adversarial settings. To address this inefficiency, we revisit the robust EMP-SSL framework, emphasizing the importance of increasing the number of crops per image to accelerate learning. Unlike traditional contrastive learning, robust EMP-SSL leverages multi-crop sampling, integrates an invariance term and regularization, and reduces training epochs, enhancing time efficiency. Evaluated with both standard linear classifiers and multi-patch embedding aggregation, robust EMP-SSL provides new insights into SSL evaluation strategies. Our results show that robust crop-based EMP-SSL not only accelerates convergence but also achieves a superior balance between clean accuracy and adversarial robustness, outperforming multi-crop embedding aggregation. Additionally, we extend this approach with free adversarial training in Multi-Crop SSL, introducing the Cost-Free Adversarial Multi-Crop Self-Supervised Learning (CF-AMC-SSL) method. CF-AMC-SSL demonstrates the effectiveness of free adversarial training in reducing training time while simultaneously improving clean accuracy and adversarial robustness. These findings underscore the potential of CF-AMC-SSL for practical SSL applications. Our code is publicly available at https://github.com/softsys4ai/CF-AMC-SSL.