Search

Large Language Models (LLMs) have shown impressive results in multiple domains of natural language processing (NLP) but are mainly focused on the English language. Recently, more LLMs have incorporated a larger proportion of multilingual text to represent low-resource languages. In Arabic NLP, several Arabic-centric LLMs have shown remarkable results on multiple benchmarks in the past two years. However, most Arabic LLMs have more than 7 billion parameters, which increases their hardware requirements and inference latency, when compared to smaller LLMs. This paper introduces Arabic Stable LM 1.6B in a base and chat version as a small but powerful Arabic-centric LLM. Our Arabic Stable LM 1.6B chat model achieves impressive results on several benchmarks beating multiple models with up to 8x the parameters. In addition, we show the benefit of mixing in synthetic instruction tuning data by augmenting our fine-tuning data with a large synthetic dialogue dataset.

Scientists and engineers use simulators to model empirically observed phenomena. However, tuning the parameters of a simulator to ensure its outputs match observed data presents a significant challenge. Simulation-based inference (SBI) addresses this by enabling Bayesian inference for simulators, identifying parameters that match observed data and align with prior knowledge. Unlike traditional Bayesian inference, SBI only needs access to simulations from the model and does not require evaluations of the likelihood-function. In addition, SBI algorithms do not require gradients through the simulator, allow for massive parallelization of simulations, and can perform inference for different observations without further simulations or training, thereby amortizing inference. Over the past years, we have developed, maintained, and extended $\texttt{sbi}$, a PyTorch-based package that implements Bayesian SBI algorithms based on neural networks. The $\texttt{sbi}$ toolkit implements a wide range of inference methods, neural network architectures, sampling methods, and diagnostic tools. In addition, it provides well-tested default settings but also offers flexibility to fully customize every step of the simulation-based inference workflow. Taken together, the $\texttt{sbi}$ toolkit enables scientists and engineers to apply state-of-the-art SBI methods to black-box simulators, opening up new possibilities for aligning simulations with empirically observed data.

Satellite imagery has dramatically revolutionized the field of geography by giving academics, scientists, and policymakers unprecedented global access to spatial data. Manual methods typically require significant time and effort to detect the generic land structure in satellite images. This study can produce a set of applications such as urban planning and development, environmental monitoring, disaster management, etc. Therefore, the research presents a methodology to minimize human labor, reducing the expenses and duration needed to identify the land structure. This article developed a deep learning-based approach to automate the process of classifying geographical land structures. We used a satellite image dataset acquired from MLRSNet. The study compared the performance of three architectures, namely CNN, ResNet-50, and Inception-v3. We used three optimizers with any model: Adam, SGD, and RMSProp. We conduct the training process for a fixed number of epochs, specifically 100 epochs, with a batch size of 64. The ResNet-50 achieved an accuracy of 76.5% with the ADAM optimizer, the Inception-v3 with RMSProp achieved an accuracy of 93.8%, and the proposed approach, CNN with RMSProp optimizer, achieved the highest level of performance and an accuracy of 94.8%. Moreover, a thorough examination of the CNN model demonstrated its exceptional accuracy, recall, and F1 scores for all categories, confirming its resilience and dependability in precisely detecting various terrain formations. The results highlight the potential of deep learning models in scene understanding, as well as their significance in efficiently identifying and categorizing land structures from satellite imagery.

During the past decade, Deep Neural Networks (DNNs) proved their value on a large variety of subjects. However despite their high value and public accessibility, the protection of the intellectual property of DNNs is still an issue and an emerging research field. Recent works have successfully extracted fully-connected DNNs using cryptanalytic methods in hard-label settings, proving that it was possible to copy a DNN with high fidelity, i.e., high similitude in the output predictions. However, the current cryptanalytic attacks cannot target complex, i.e., not fully connected, DNNs and are limited to special cases of neurons present in deep networks. In this work, we introduce a new end-to-end attack framework designed for model extraction of embedded DNNs with high fidelity. We describe a new black-box side-channel attack which splits the DNN in several linear parts for which we can perform cryptanalytic extraction and retrieve the weights in hard-label settings. With this method, we are able to adapt cryptanalytic extraction, for the first time, to non-fully connected DNNs, while maintaining a high fidelity. We validate our contributions by targeting several architectures implemented on a microcontroller unit, including a Multi-Layer Perceptron (MLP) of 1.7 million parameters and a shortened MobileNetv1. Our framework successfully extracts all of these DNNs with high fidelity (88.4% for the MobileNetv1 and 93.2% for the MLP). Furthermore, we use the stolen model to generate adversarial examples and achieve close to white-box performance on the victim's model (95.8% and 96.7% transfer rate).

All-dielectric metasurfaces can produce structural colors, but the most advantageous design criteria are still being investigated. This work numerically studies how the two-dimensional shape of nanoparticles affects the colorimetric response under circularly polarized light (CPL) to develop a sensor distinguishing CPL orientations. Using lossless dielectric materials (silicon nitride on silicon dioxide), we achieve far-field dichroism by modifying oblong nanoparticles into L-shaped structures through corner cuts. This design suppresses one electric dipole under CPL illumination, leading to differential colorimetric responses. We link these responses to a decoupling effect in the near-field net electric flux. Our findings provide design guidelines for all-dielectric, lossless colorimetric sensors of chiral light., Comment: This is an unedited version of a work submitted to Optics Express, with 46 pages (15 main text, 31 supplemental document), 28 figures (5 main text, 23 supplemental document), and 8 tables (all supplemental document). Dataset and Code files are with the Optics Express submission

Surgical scene segmentation is essential for enhancing surgical precision, yet it is frequently compromised by the scarcity and imbalance of available data. To address these challenges, semantic image synthesis methods based on generative adversarial networks and diffusion models have been developed. However, these models often yield non-diverse images and fail to capture small, critical tissue classes, limiting their effectiveness. In response, we propose the Class-Aware Semantic Diffusion Model (CASDM), a novel approach which utilizes segmentation maps as conditions for image synthesis to tackle data scarcity and imbalance. Novel class-aware mean squared error and class-aware self-perceptual loss functions have been defined to prioritize critical, less visible classes, thereby enhancing image quality and relevance. Furthermore, to our knowledge, we are the first to generate multi-class segmentation maps using text prompts in a novel fashion to specify their contents. These maps are then used by CASDM to generate surgical scene images, enhancing datasets for training and validating segmentation models. Our evaluation, which assesses both image quality and downstream segmentation performance, demonstrates the strong effectiveness and generalisability of CASDM in producing realistic image-map pairs, significantly advancing surgical scene segmentation across diverse and challenging datasets.

Recently published work on rephrasing natural text data for pre-training LLMs has shown promising results when combining the original dataset with the synthetically rephrased data. We build upon previous work by replicating existing results on C4 and extending them with our optimized rephrasing pipeline to the English, German, Italian, and Spanish Oscar subsets of CulturaX. Our pipeline leads to increased performance on standard evaluation benchmarks in both the mono- and multilingual setup. In addition, we provide a detailed study of our pipeline, investigating the choice of the base dataset and LLM for the rephrasing, as well as the relationship between the model size and the performance after pre-training. By exploring data with different perceived quality levels, we show that gains decrease with higher quality. Furthermore, we find the difference in performance between model families to be bigger than between different model sizes. This highlights the necessity for detailed tests before choosing an LLM to rephrase large amounts of data. Moreover, we investigate the effect of pre-training with synthetic data on supervised fine-tuning. Here, we find increasing but inconclusive results that highly depend on the used benchmark. These results (again) highlight the need for better benchmarking setups. In summary, we show that rephrasing multilingual and low-quality data is a very promising direction to extend LLM pre-training data., Comment: 21 pages, 4 figures, 12 tables

The rise of social media and short-form video (SFV) has facilitated a breeding ground for misinformation. With the emergence of large language models, significant research has gone into curbing this misinformation problem with automatic false claim detection for text. Unfortunately, the automatic detection of misinformation in SFV is a more complex problem that remains largely unstudied. While text samples are monomodal (only containing words), SFVs comprise three different modalities: words, visuals, and non-linguistic audio. In this work, we introduce Video Masked Autoencoders for Misinformation Guarding (ViMGuard), the first deep-learning architecture capable of fact-checking an SFV through analysis of all three of its constituent modalities. ViMGuard leverages a dual-component system. First, Video and Audio Masked Autoencoders analyze the visual and non-linguistic audio elements of a video to discern its intention; specifically whether it intends to make an informative claim. If it is deemed that the SFV has informative intent, it is passed through our second component: a Retrieval Augmented Generation system that validates the factual accuracy of spoken words. In evaluation, ViMGuard outperformed three cutting-edge fact-checkers, thus setting a new standard for SFV fact-checking and marking a significant stride toward trustworthy news on social platforms. To promote further testing and iteration, VimGuard was deployed into a Chrome extension and all code was open-sourced on GitHub., Comment: 7 pages, 2 figures

Evidence of chirality was observed at the Fe metal center in Fe(III) spin crossover coordination salts [Fe(qsal)2][ Ni(dmit)2] and [Fe(qsal)2](TCNQ)2 from X-ray absorption spectroscopy at the Fe 2p3/2 core threshold. This indicates the formation of chiral domains that influence the octahedral coordination on the Fe core., Comment: 5 pages, 4 figures. Submitted for publication in JPCM

The process of creating training data to teach models is currently driven by humans, who manually analyze model weaknesses and plan how to create data that improves a student model. Recent approaches using LLMs as annotators reduce human effort, but still require humans to interpret feedback from evaluations and control the LLM to produce data the student needs. Automating this labor-intensive process by creating autonomous data generation agents - or teachers - is desirable, but requires environments that can simulate the feedback-driven, iterative, closed loop of data creation. To enable rapid and scalable testing for such agents and their modules, we introduce DataEnvGym, a testbed of teacher environments for data generation agents. DataEnvGym frames data generation as a sequential decision-making task, involving an agent consisting of a data generation policy (which generates a plan for creating training data) and a data generation engine (which transforms the plan into data), inside an environment that provides student feedback. The agent's goal is to improve student performance. Students are iteratively trained and evaluated on generated data, with their feedback (in the form of errors or weak skills) being reported to the agent after each iteration. DataEnvGym includes multiple teacher environment instantiations across 3 levels of structure in the state representation and action space. More structured environments are based on inferred skills and offer more interpretability and curriculum control. We support 3 diverse tasks (math, code, and VQA) and test multiple students and teachers. Example agents in our teaching environments can iteratively improve students across tasks and settings. Moreover, we show that environments teach different skill levels and test variants of key modules, pointing to future work in improving data generation agents, engines, and feedback mechanisms., Comment: Project Page: https://DataEnvGym.github.io

Image restoration and spectral reconstruction are longstanding computer vision tasks. Currently, CNN-transformer hybrid models provide state-of-the-art performance for these tasks. The key common ingredient in the architectural designs of these models is Channel-wise Self-Attention (CSA). We first show that CSA is an overall low-rank operation. Then, we propose an instance-Guided Low-rank Multi-Head selfattention (GLMHA) to replace the CSA for a considerable computational gain while closely retaining the original model performance. Unique to the proposed GLMHA is its ability to provide computational gain for both short and long input sequences. In particular, the gain is in terms of both Floating Point Operations (FLOPs) and parameter count reduction. This is in contrast to the existing popular computational complexity reduction techniques, e.g., Linformer, Performer, and Reformer, for whom FLOPs overpower the efficient design tricks for the shorter input sequences. Moreover, parameter reduction remains unaccounted for in the existing methods.We perform an extensive evaluation for the tasks of spectral reconstruction from RGB images, spectral reconstruction from snapshot compressive imaging, motion deblurring, and image deraining by enhancing the best-performing models with our GLMHA. Our results show up to a 7.7 Giga FLOPs reduction with 370K fewer parameters required to closely retain the original performance of the best-performing models that employ CSA.

Crack detection, particularly from pavement images, presents a formidable challenge in the domain of computer vision due to several inherent complexities such as intensity inhomogeneity, intricate topologies, low contrast, and noisy backgrounds. Automated crack detection is crucial for maintaining the structural integrity of essential infrastructures, including buildings, pavements, and bridges. Existing lightweight methods often face challenges including computational inefficiency, complex crack patterns, and difficult backgrounds, leading to inaccurate detection and impracticality for real-world applications. To address these limitations, we propose EfficientCrackNet, a lightweight hybrid model combining Convolutional Neural Networks (CNNs) and transformers for precise crack segmentation. EfficientCrackNet integrates depthwise separable convolutions (DSC) layers and MobileViT block to capture both global and local features. The model employs an Edge Extraction Method (EEM) and for efficient crack edge detection without pretraining, and Ultra-Lightweight Subspace Attention Module (ULSAM) to enhance feature extraction. Extensive experiments on three benchmark datasets Crack500, DeepCrack, and GAPs384 demonstrate that EfficientCrackNet achieves superior performance compared to existing lightweight models, while requiring only 0.26M parameters, and 0.483 FLOPs (G). The proposed model offers an optimal balance between accuracy and computational efficiency, outperforming state-of-the-art lightweight models, and providing a robust and adaptable solution for real-world crack segmentation.

The proliferation of Large Language Models (LLMs) in diverse applications underscores the pressing need for robust security measures to thwart potential jailbreak attacks. These attacks exploit vulnerabilities within LLMs, endanger data integrity and user privacy. Guardrails serve as crucial protective mechanisms against such threats, but existing models often fall short in terms of both detection accuracy, and computational efficiency. This paper advocates for the significance of jailbreak attack prevention on LLMs, and emphasises the role of input guardrails in safeguarding these models. We introduce MoJE (Mixture of Jailbreak Expert), a novel guardrail architecture designed to surpass current limitations in existing state-of-the-art guardrails. By employing simple linguistic statistical techniques, MoJE excels in detecting jailbreak attacks while maintaining minimal computational overhead during model inference. Through rigorous experimentation, MoJE demonstrates superior performance capable of detecting 90% of the attacks without compromising benign prompts, enhancing LLMs security against jailbreak attacks.

Cardiovascular Diseases (CVDs) are the leading cause of death worldwide, taking 17.9 million lives annually. Abdominal Aortic Calcification (AAC) is an established marker for CVD, which can be observed in lateral view Vertebral Fracture Assessment (VFA) scans, usually done for vertebral fracture detection. Early detection of AAC may help reduce the risk of developing clinical CVDs by encouraging preventive measures. Manual analysis of VFA scans for AAC measurement is time consuming and requires trained human assessors. Recently, efforts have been made to automate the process, however, the proposed models are either low in accuracy, lack granular level score prediction, or are too heavy in terms of inference time and memory footprint. Considering all these shortcomings of existing algorithms, we propose 'AACLiteNet', a lightweight deep learning model that predicts both cumulative and granular level AAC scores with high accuracy, and also has a low memory footprint, and computation cost (Floating Point Operations (FLOPs)). The AACLiteNet achieves a significantly improved one-vs-rest average accuracy of 85.94% as compared to the previous best 81.98%, with 19.88 times less computational cost and 2.26 times less memory footprint, making it implementable on portable computing devices., Comment: 10 pages including references

As generative AI, particularly large language models (LLMs), become increasingly integrated into production applications, new attack surfaces and vulnerabilities emerge and put a focus on adversarial threats in natural language and multi-modal systems. Red-teaming has gained importance in proactively identifying weaknesses in these systems, while blue-teaming works to protect against such adversarial attacks. Despite growing academic interest in adversarial risks for generative AI, there is limited guidance tailored for practitioners to assess and mitigate these challenges in real-world environments. To address this, our contributions include: (1) a practical examination of red- and blue-teaming strategies for securing generative AI, (2) identification of key challenges and open questions in defense development and evaluation, and (3) the Attack Atlas, an intuitive framework that brings a practical approach to analyzing single-turn input attacks, placing it at the forefront for practitioners. This work aims to bridge the gap between academic insights and practical security measures for the protection of generative AI systems.

This paper introduces TINA, a novel framework for implementing non Neural Network (NN) signal processing algorithms on NN accelerators such as GPUs, TPUs or FPGAs. The key to this approach is the concept of mapping mathematical and logic functions as a series of convolutional and fully connected layers. By mapping functions into such a small substack of NN layers, it becomes possible to execute non-NN algorithms on NN hardware (HW) accelerators efficiently, as well as to ensure the portability of TINA implementations to any platform that supports such NN accelerators. Results show that TINA is highly competitive compared to alternative frameworks, specifically for complex functions with iterations. For a Polyphase Filter Bank use case TINA shows GPU speedups of up to 80x vs a CPU baseline with NumPy compared to 8x speedup achieved by alternative frameworks. The framework is open source and publicly available at https://github.com/ChristiaanBoe/TINA., Comment: Preprint for MLSP 2024

Authorship obfuscation, rewriting a text to intentionally obscure the identity of the author, is an important but challenging task. Current methods using large language models (LLMs) lack interpretability and controllability, often ignoring author-specific stylistic features, resulting in less robust performance overall. To address this, we develop StyleRemix, an adaptive and interpretable obfuscation method that perturbs specific, fine-grained style elements of the original input text. StyleRemix uses pre-trained Low Rank Adaptation (LoRA) modules to rewrite an input specifically along various stylistic axes (e.g., formality and length) while maintaining low computational cost. StyleRemix outperforms state-of-the-art baselines and much larger LLMs in a variety of domains as assessed by both automatic and human evaluation. Additionally, we release AuthorMix, a large set of 30K high-quality, long-form texts from a diverse set of 14 authors and 4 domains, and DiSC, a parallel corpus of 1,500 texts spanning seven style axes in 16 unique directions

Data balancing across multiple modalities/sources appears in various forms in several foundation models (e.g., CLIP and DINO) achieving universal representation learning. We show that this iterative algorithm, usually used to avoid representation collapse, enjoys an unsuspected benefit: reducing the variance of estimators that are functionals of the empirical distribution over these sources. We provide non-asymptotic bounds quantifying this variance reduction effect and relate them to the eigendecays of appropriately defined Markov operators. We explain how various forms of data balancing in contrastive multimodal learning and self-supervised clustering can be interpreted as instances of this variance reduction scheme.

As urban air mobility (UAM) emerges as a transformative solution to urban transportation, the demand for robust communication frameworks capable of supporting high-density aerial traffic becomes increasingly critical. An essential area of communications improvement is reliably characterizing and minimizing interference on UAM aircraft from other aircraft and ground vehicles. To achieve this, accurate line-of-sight (LOS) models must be used. In this work, we highlight the limitations of a LOS probability model extensively used in the literature in accurately predicting interference caused by ground vehicles. Then, we introduce a modified probability of LOS model that improves interference prediction by incorporating the urban topography and the dynamic positioning of ground vehicles on streets. Our model's parameters are derived from extensive simulations and validated through real-world urban settings to ensure reliability and applicability.

Chisel (Constructing Hardware In a Scala Embedded Language) is a broadly adopted HDL that brings object-oriented and functional programming, type-safety, and parameterization to hardware design. However, while these language features significantly improve the process of writing code, debugging Chisel designs with open source tools loses many of the advantages of the source language, as type information and data structure hierarchies are lost in the translation, simulator output, and waveform viewer. This work, Tywaves, presents a new type-centered debugging format that brings the same level of abstraction found in contemporary hardware languages to waveform viewers. Contributions to the Chisel library and CIRCT MLIR compiler as well as the Surfer waveform viewer result in a waveform viewer that better supports the Chisel HDL. Project url: https://github.com/rameloni/tywaves-chisel-demo, Comment: 6 pages, 2 tables, 3 figures

From a comparison of energy dispersive analyses of X-rays (EDAX), the known molecular stoichiometry and X-ray photoemission spectroscopy (XPS), it is evident that the Fe(III) spin crossover salt [Fe(qsal)2Ni(dmit)2] has a preferential surface termination with the Ni(dmit)2 moiety. This preferential surface termination leads to a significant surface to bulk core level shift for the Ni 2p X-ray photoemission core level, not seen in the corresponding Fe 2p core level spectra. Inverse photoemission spectroscopy (IPES), thus provides some indication of the density of states resulting from the dmit2 = 1,3-dithiol-2-thione-4,5-dithiolato ligand unoccupied molecular orbitals., Comment: 5 pages, 4 figures, submitted for publication in JPCM

Customers reach out to online live chat agents with various intents, such as asking about product details or requesting a return. In this paper, we propose the problem of predicting user intent from browsing history and address it through a two-stage approach. The first stage classifies a user's browsing history into high-level intent categories. Here, we represent each browsing history as a text sequence of page attributes and use the ground-truth class labels to fine-tune pretrained Transformers. The second stage provides a large language model (LLM) with the browsing history and predicted intent class to generate fine-grained intents. For automatic evaluation, we use a separate LLM to judge the similarity between generated and ground-truth intents, which closely aligns with human judgments. Our two-stage approach yields significant performance gains compared to generating intents without the classification stage., Comment: CIKM 2024

In this paper, we use open-source tools to perform quantum resource estimation to assess the requirements for industry-relevant quantum computation. Our analysis uses the problem of industrial shift scheduling in manufacturing and the Quantum Industrial Shift Scheduling algorithm. We base our figures of merit on current technology, as well as theoretical high-fidelity scenarios for superconducting qubit platforms. We find that the execution time of gate and measurement operations determines the overall computational runtime more strongly than the system error rates. Moreover, achieving a quantum speedup would not only require low system error rates ($10^{-6}$ or better), but also measurement operations with an execution time below 10ns. This rules out the possibility of near-term quantum advantage for this use case, and suggests that significant technological or algorithmic progress will be needed before such an advantage can be achieved.

In this data article, we introduce the Multi-Modal Event-based Vehicle Detection and Tracking (MEVDT) dataset. This dataset provides a synchronized stream of event data and grayscale images of traffic scenes, captured using the Dynamic and Active-Pixel Vision Sensor (DAVIS) 240c hybrid event-based camera. MEVDT comprises 63 multi-modal sequences with approximately 13k images, 5M events, 10k object labels, and 85 unique object tracking trajectories. Additionally, MEVDT includes manually annotated ground truth labels $\unicode{x2014}$ consisting of object classifications, pixel-precise bounding boxes, and unique object IDs $\unicode{x2014}$ which are provided at a labeling frequency of 24 Hz. Designed to advance the research in the domain of event-based vision, MEVDT aims to address the critical need for high-quality, real-world annotated datasets that enable the development and evaluation of object detection and tracking algorithms in automotive environments.

We present foundation language models developed to power Apple Intelligence features, including a ~3 billion parameter model designed to run efficiently on devices and a large server-based language model designed for Private Cloud Compute. These models are designed to perform a wide range of tasks efficiently, accurately, and responsibly. This report describes the model architecture, the data used to train the model, the training process, how the models are optimized for inference, and the evaluation results. We highlight our focus on Responsible AI and how the principles are applied throughout the model development.

Graph Neural Networks (GNNs) are widely used today in recommendation systems, fraud detection, and node/link classification tasks. Real world GNNs continue to scale in size and require a large memory footprint for storing graphs and embeddings that often exceed the memory capacities of the target GPUs used for training. To address limited memory capacities, traditional GNN training approaches use graph partitioning and sharding techniques to scale up across multiple GPUs within a node and/or scale out across multiple nodes. However, this approach suffers from the high computational costs of graph partitioning algorithms and inefficient communication across GPUs. To address these overheads, we propose Large-scale Storage-based Multi-GPU GNN framework (LSM-GNN), a storagebased approach to train GNN models that utilizes a novel communication layer enabling GPU software caches to function as a system-wide shared cache with low overheads.LSM-GNN incorporates a hybrid eviction policy that intelligently manages cache space by using both static and dynamic node information to significantly enhance cache performance. Furthermore, we introduce the Preemptive Victim-buffer Prefetcher (PVP), a mechanism for prefetching node feature data from a Victim Buffer located in CPU pinned-memory to further reduce the pressure on the storage devices. Experimental results show that despite the lower compute capabilities and memory capacities, LSM-GNN in a single node with two GPUs offers superior performance over two-node-four-GPU Dist-DGL baseline and provides up to 3.75x speed up on end-to-end epoch time while running large-scale GNN training

Spin crossover (SCO) complexes are highly promising candidates for a myriad of potential applications in room-temperature electronics; however, as it stands, establishing a clear connection between their spin-state switching and transport properties has been far from trivial. In this letter, an effort to unravel the underlying charge transport mechanism in these SCO complexes, via a general theory, is made. The theory presented herein is aimed at providing a unifying picture that explains the widely different trends observed in the spin-crossover-dependent carrier transport properties in the SCO molecular thin film systems.

Robust perception is critical for autonomous driving, especially under adverse weather and lighting conditions that commonly occur in real-world environments. In this paper, we introduce the Stereo Image Dataset (SID), a large-scale stereo-image dataset that captures a wide spectrum of challenging real-world environmental scenarios. Recorded at a rate of 20 Hz using a ZED stereo camera mounted on a vehicle, SID consists of 27 sequences totaling over 178k stereo image pairs that showcase conditions from clear skies to heavy snow, captured during the day, dusk, and night. The dataset includes detailed sequence-level annotations for weather conditions, time of day, location, and road conditions, along with instances of camera lens soiling, offering a realistic representation of the challenges in autonomous navigation. Our work aims to address a notable gap in research for autonomous driving systems by presenting high-fidelity stereo images essential for the development and testing of advanced perception algorithms. These algorithms support consistent and reliable operation across variable weather and lighting conditions, even when handling challenging situations like lens soiling. SID is publicly available at: https://doi.org/10.7302/esz6-nv83., Comment: This paper has been accepted for publication at the 2024 IEEE National Aerospace and Electronics Conference (NAECON), 2024

When training deep learning models, the performance depends largely on the selected hyperparameters. However, hyperparameter optimization (HPO) is often one of the most expensive parts of model design. Classical HPO methods treat this as a black-box optimization problem. However, gray-box HPO methods, which incorporate more information about the setup, have emerged as a promising direction for more efficient optimization. For example, using intermediate loss evaluations to terminate bad selections. In this work, we propose an HPO method for neural networks using logged checkpoints of the trained weights to guide future hyperparameter selections. Our method, Forecasting Model Search (FMS), embeds weights into a Gaussian process deep kernel surrogate model, using a permutation-invariant graph metanetwork to be data-efficient with the logged network weights. To facilitate reproducibility and further research, we open-source our code at https://github.com/NVlabs/forecasting-model-search., Comment: See the project website at https://research.nvidia.com/labs/toronto-ai/FMS/

One of the most striking findings in modern research on large language models (LLMs) is that scaling up compute during training leads to better results. However, less attention has been given to the benefits of scaling compute during inference. This survey focuses on these inference-time approaches. We explore three areas under a unified mathematical formalism: token-level generation algorithms, meta-generation algorithms, and efficient generation. Token-level generation algorithms, often called decoding algorithms, operate by sampling a single token at a time or constructing a token-level search space and then selecting an output. These methods typically assume access to a language model's logits, next-token distributions, or probability scores. Meta-generation algorithms work on partial or full sequences, incorporating domain knowledge, enabling backtracking, and integrating external information. Efficient generation methods aim to reduce token costs and improve the speed of generation. Our survey unifies perspectives from three research communities: traditional natural language processing, modern LLMs, and machine learning systems.

The terahertz communications have the potential to revolutionize data transfer with unmatched speed and facilitate the development of new high-bandwidth applications. This paper studies the performance of downlink terahertz system assisted by beyond diagonal intelligent reconfigurable surface (BD-IRS). For enhanced energy efficiency and low cost, a joint precoding and BD-IRS phase shift design satisfying the $1$-bit resolution constraints to maximize the spectral efficiency is presented. The original problem is non-linear, NP-hard, and intricately coupled, and obtaining an optimal solution is challenging. To reduce the complexity, we first transform the optimization problem into two problems and then iteratively solve them to achieve an efficient solution. Numerical results demonstrate that the proposed approach for the BD-IRS assisted terahertz system significantly enhances the spectral efficiency compared to the conventional diagonal IRS assisted system., Comment: 5,3

Formal verification has emerged as a promising method to ensure the safety and reliability of neural networks. Naively verifying a safety property amounts to ensuring the safety of a neural network for the whole input space irrespective of any training or test set. However, this also implies that the safety of the neural network is checked even for inputs that do not occur in the real-world and have no meaning at all, often resulting in spurious errors. To tackle this shortcoming, we propose the VeriFlow architecture as a flow based density model tailored to allow any verification approach to restrict its search to the some data distribution of interest. We argue that our architecture is particularly well suited for this purpose because of two major properties. First, we show that the transformation and log-density function that are defined by our model are piece-wise affine. Therefore, the model allows the usage of verifiers based on SMT with linear arithmetic. Second, upper density level sets (UDL) of the data distribution take the shape of an $L^p$-ball in the latent space. As a consequence, representations of UDLs specified by a given probability are effectively computable in latent space. This allows for SMT and abstract interpretation approaches with fine-grained, probabilistically interpretable, control regarding on how (a)typical the inputs subject to verification are.

We introduce a novel discretization scheme for Wasserstein gradient flows that involves successively computing Schr\"{o}dinger bridges with the same marginals. This is different from both the forward/geodesic approximation and the backward/Jordan-Kinderlehrer-Otto (JKO) approximations. The proposed scheme has two advantages: one, it avoids the use of the score function, and, two, it is amenable to particle-based approximations using the Sinkhorn algorithm. Our proof hinges upon showing that relative entropy between the Schr\"{o}dinger bridge with the same marginals at temperature $\epsilon$ and the joint distribution of a stationary Langevin diffusion at times zero and $\epsilon$ is of the order $o(\epsilon^2)$ with an explicit dependence given by Fisher information. Owing to this inequality, we can show, using a triangular approximation argument, that the interpolated iterated application of the Schr\"{o}dinger bridge approximation converge to the Wasserstein gradient flow, for a class of gradient flows, including the heat flow. The results also provide a probabilistic and rigorous framework for the convergence of the self-attention mechanisms in transformer networks to the solutions of heat flows, first observed in the inspiring work SABP22 in machine learning research., Comment: 36 pages, 1 figure

Over 85 million computed tomography (CT) scans are performed annually in the US, of which approximately one quarter focus on the abdomen. Given the current radiologist shortage, there is a large impetus to use artificial intelligence to alleviate the burden of interpreting these complex imaging studies. Prior state-of-the-art approaches for automated medical image interpretation leverage vision language models (VLMs). However, current medical VLMs are generally limited to 2D images and short reports, and do not leverage electronic health record (EHR) data for supervision. We introduce Merlin - a 3D VLM that we train using paired CT scans (6+ million images from 15,331 CTs), EHR diagnosis codes (1.8+ million codes), and radiology reports (6+ million tokens). We evaluate Merlin on 6 task types and 752 individual tasks. The non-adapted (off-the-shelf) tasks include zero-shot findings classification (31 findings), phenotype classification (692 phenotypes), and zero-shot cross-modal retrieval (image to findings and image to impressions), while model adapted tasks include 5-year disease prediction (6 diseases), radiology report generation, and 3D semantic segmentation (20 organs). We perform internal validation on a test set of 5,137 CTs, and external validation on 7,000 clinical CTs and on two public CT datasets (VerSe, TotalSegmentator). Beyond these clinically-relevant evaluations, we assess the efficacy of various network architectures and training strategies to depict that Merlin has favorable performance to existing task-specific baselines. We derive data scaling laws to empirically assess training data needs for requisite downstream task performance. Furthermore, unlike conventional VLMs that require hundreds of GPUs for training, we perform all training on a single GPU., Comment: 18 pages, 7 figures

This paper presents a new boundary-value problem formulation for quantifying uncertainty induced by the presence of small Brownian noise near transversally stable periodic orbits (limit cycles) and quasiperiodic invariant tori of the deterministic dynamical systems obtained in the absence of noise. The formulation uses adjoints to construct a continuous family of transversal hyperplanes that are invariant under the linearized deterministic flow near the limit cycle or quasiperiodic invariant torus. The intersections with each hyperplane of stochastic trajectories that remain near the deterministic cycle or torus over intermediate times may be approximated by a Gaussian distribution whose covariance matrix can be obtained from the solution to the corresponding boundary-value problem. In the case of limit cycles, the analysis improves upon results in the literature through the explicit use of state-space projections, transversality constraints, and symmetry-breaking parameters that ensure uniqueness of the solution despite the lack of hyperbolicity along the limit cycle. These same innovations are then generalized to the case of a quasiperiodic invariant torus of arbitrary dimension. In each case, a closed-form solution to the covariance boundary-value problem is found in terms of a convergent series. The methodology is validated against the results of numerical integration for two examples of stochastically perturbed limit cycles and one example of a stochastically perturbed two-dimensional quasiperiodic invariant torus. Finally, an implementation of the covariance boundary-value problem in the numerical continuation package coco is applied to analyze the small-noise limit near a two-dimensional quasiperiodic invariant torus in a nonlinear deterministic dynamical system in $\mathbb{R}^4$ that does not support closed-form analysis., Comment: abstract had to be shortened due to 1920 character limitation of arXiv

Many neurological conditions, e.g., a stroke, can cause patients to experience upper limb (UL) motor impairments that hinder their daily activities. For such patients, while rehabilitation therapy is key for regaining autonomy and restoring mobility, its long-term nature entails ongoing time commitment and it is often not sufficiently engaging. Virtual reality (VR) can transform rehabilitation therapy into engaging game-like tasks that can be tailored to patient-specific activities, set goals, and provide rehabilitation assessment. Yet, most VR systems lack built-in methods to track progress over time and alter rehabilitation programs accordingly. We propose using arm kinematic modeling and capability maps to allow a VR system to understand a user's physical capability and limitation. Next, we suggest two use cases for the VR system to utilize the user's capability map for tailoring rehabilitation programs. Finally, for one use case, it is shown that the VR system can emphasize and assess the use of specific UL joints., Comment: 4 pages, 4 figures. Paper accepted at IEEE International Conference on Engineering in Medicine & Biology Society, 2024

Several therapy routines require deep breathing exercises as a key component and patients undergoing such therapies must perform these exercises regularly. Assessing the outcome of a therapy and tailoring its course necessitates monitoring a patient's compliance with the therapy. While therapy compliance monitoring is routine in a clinical environment, it is challenging to do in an at-home setting. This is so because a home setting lacks access to specialized equipment and skilled professionals needed to effectively monitor the performance of a therapy routine by a patient. For some types of therapies, these challenges can be addressed with the use of consumer-grade hardware, such as earphones and smartphones, as practical solutions. To accurately monitor breathing exercises using wireless earphones, this paper proposes a framework that has the potential for assessing a patient's compliance with an at-home therapy. The proposed system performs real-time detection of breathing phases and channels with high accuracy by processing a $\mathbf{500}$ ms audio signal through two convolutional neural networks. The first network, called a channel classifier, distinguishes between nasal and oral breathing, and a pause. The second network, called a phase classifier, determines whether the audio segment is from inhalation or exhalation. According to $k$-fold cross-validation, the channel and phase classifiers achieved a maximum F1 score of $\mathbf{97.99\%}$ and $\mathbf{89.46\%}$, respectively. The results demonstrate the potential of using commodity earphones for real-time breathing channel and phase detection for breathing therapy compliance monitoring., Comment: 4 pages, 2 figures. Paper accepted at IEEE International Conference on Engineering in Medicine & Biology Society, 2024

The goal of selective prediction is to allow an a model to abstain when it may not be able to deliver a reliable prediction, which is important in safety-critical contexts. Existing approaches to selective prediction typically require access to the internals of a model, require retraining a model or study only unimodal models. However, the most powerful models (e.g. GPT-4) are typically only available as black boxes with inaccessible internals, are not retrainable by end-users, and are frequently used for multimodal tasks. We study the possibility of selective prediction for vision-language models in a realistic, black-box setting. We propose using the principle of \textit{neighborhood consistency} to identify unreliable responses from a black-box vision-language model in question answering tasks. We hypothesize that given only a visual question and model response, the consistency of the model's responses over the neighborhood of a visual question will indicate reliability. It is impossible to directly sample neighbors in feature space in a black-box setting. Instead, we show that it is possible to use a smaller proxy model to approximately sample from the neighborhood. We find that neighborhood consistency can be used to identify model responses to visual questions that are likely unreliable, even in adversarial settings or settings that are out-of-distribution to the proxy model., Comment: CVPR 2024