Search

To restore proper blood flow in blocked coronary arteries via angioplasty procedure, accurate placement of devices such as catheters, balloons, and stents under live fluoroscopy or diagnostic angiography is crucial. Identified balloon markers help in enhancing stent visibility in X-ray sequences, while the catheter tip aids in precise navigation and co-registering vessel structures, reducing the need for contrast in angiography. However, accurate detection of these devices in interventional X-ray sequences faces significant challenges, particularly due to occlusions from contrasted vessels and other devices and distractions from surrounding, resulting in the failure to track such small objects. While most tracking methods rely on spatial correlation of past and current appearance, they often lack strong motion comprehension essential for navigating through these challenging conditions, and fail to effectively detect multiple instances in the scene. To overcome these limitations, we propose a self-supervised learning approach that enhances its spatio-temporal understanding by incorporating supplementary cues and learning across multiple representation spaces on a large dataset. Followed by that, we introduce a generic real-time tracking framework that effectively leverages the pretrained spatio-temporal network and also takes the historical appearance and trajectory data into account. This results in enhanced localization of multiple instances of device landmarks. Our method outperforms state-of-the-art methods in interventional X-ray device tracking, especially stability and robustness, achieving an 87% reduction in max error for balloon marker detection and a 61% reduction in max error for catheter tip detection.

Chart visualizations, while essential for data interpretation and communication, are predominantly accessible only as images in PDFs, lacking source data tables and stylistic information. To enable effective editing of charts in PDFs or digital scans, we present PlotEdit, a novel multi-agent framework for natural language-driven end-to-end chart image editing via self-reflective LLM agents. PlotEdit orchestrates five LLM agents: (1) Chart2Table for data table extraction, (2) Chart2Vision for style attribute identification, (3) Chart2Code for retrieving rendering code, (4) Instruction Decomposition Agent for parsing user requests into executable steps, and (5) Multimodal Editing Agent for implementing nuanced chart component modifications - all coordinated through multimodal feedback to maintain visual fidelity. PlotEdit outperforms existing baselines on the ChartCraft dataset across style, layout, format, and data-centric edits, enhancing accessibility for visually challenged users and improving novice productivity., Comment: Accepted at ECIR 2025

Prospective students face the challenging task of selecting a university program that will shape their academic and professional careers. For decision-makers and support services, it is often time-consuming and extremely difficult to match personal interests with suitable programs due to the vast and complex catalogue information available. This paper presents the first information system that provides students with efficient recommendations based on both program content and personal preferences. BERTopic, a powerful topic modeling algorithm, is used that leverages text embedding techniques to generate topic representations. It enables us to mine interest topics from all course descriptions, representing the full body of knowledge taught at the institution. Underpinned by the student's individual choice of topics, a shortlist of the most relevant programs is computed through statistical backtracking in the knowledge map, a novel characterization of the program-course relationship. This approach can be applied to a wide range of educational settings, including professional and vocational training. A case study at a post-secondary school with 80 programs and over 5,000 courses shows that the system provides immediate and effective decision support. The presented interest topics are meaningful, leading to positive effects such as serendipity, personalization, and fairness, as revealed by a qualitative study involving 65 students. Over 98% of users indicated that the recommendations aligned with their interests, and about 94% stated they would use the tool in the future. Quantitative analysis shows the system can be configured to ensure fairness, achieving 98% program coverage while maintaining a personalization score of 0.77. These findings suggest that this real-time, user-centered, data-driven system could improve the program selection process., Comment: Accepted at Data Mining and Knowledge Discovery (Springer)

Safety-critical applications, such as autonomous driving, require extensive multimodal data for rigorous testing. Methods based on synthetic data are gaining prominence due to the cost and complexity of gathering real-world data but require a high degree of realism and controllability in order to be useful. This paper introduces MObI, a novel framework for Multimodal Object Inpainting that leverages a diffusion model to create realistic and controllable object inpaintings across perceptual modalities, demonstrated for both camera and lidar simultaneously. Using a single reference RGB image, MObI enables objects to be seamlessly inserted into existing multimodal scenes at a 3D location specified by a bounding box, while maintaining semantic consistency and multimodal coherence. Unlike traditional inpainting methods that rely solely on edit masks, our 3D bounding box conditioning gives objects accurate spatial positioning and realistic scaling. As a result, our approach can be used to insert novel objects flexibly into multimodal scenes, providing significant advantages for testing perception models., Comment: 8 pages

In this work, we propose a Multi-LLM summarization framework, and investigate two different multi-LLM strategies including centralized and decentralized. Our multi-LLM summarization framework has two fundamentally important steps at each round of conversation: generation and evaluation. These steps are different depending on whether our multi-LLM decentralized summarization is used or centralized. In both our multi-LLM decentralized and centralized strategies, we have k different LLMs that generate diverse summaries of the text. However, during evaluation, our multi-LLM centralized summarization approach leverages a single LLM to evaluate the summaries and select the best one whereas k LLMs are used for decentralized multi-LLM summarization. Overall, we find that our multi-LLM summarization approaches significantly outperform the baselines that leverage only a single LLM by up to 3x. These results indicate the effectiveness of multi-LLM approaches for summarization.

Graphical User Interface (GUI) agents, powered by Large Foundation Models, have emerged as a transformative approach to automating human-computer interaction. These agents autonomously interact with digital systems or software applications via GUIs, emulating human actions such as clicking, typing, and navigating visual elements across diverse platforms. Motivated by the growing interest and fundamental importance of GUI agents, we provide a comprehensive survey that categorizes their benchmarks, evaluation metrics, architectures, and training methods. We propose a unified framework that delineates their perception, reasoning, planning, and acting capabilities. Furthermore, we identify important open challenges and discuss key future directions. Finally, this work serves as a basis for practitioners and researchers to gain an intuitive understanding of current progress, techniques, benchmarks, and critical open problems that remain to be addressed.

Understanding information from a collection of multiple documents, particularly those with visually rich elements, is important for document-grounded question answering. This paper introduces VisDoMBench, the first comprehensive benchmark designed to evaluate QA systems in multi-document settings with rich multimodal content, including tables, charts, and presentation slides. We propose VisDoMRAG, a novel multimodal Retrieval Augmented Generation (RAG) approach that simultaneously utilizes visual and textual RAG, combining robust visual retrieval capabilities with sophisticated linguistic reasoning. VisDoMRAG employs a multi-step reasoning process encompassing evidence curation and chain-of-thought reasoning for concurrent textual and visual RAG pipelines. A key novelty of VisDoMRAG is its consistency-constrained modality fusion mechanism, which aligns the reasoning processes across modalities at inference time to produce a coherent final answer. This leads to enhanced accuracy in scenarios where critical information is distributed across modalities and improved answer verifiability through implicit context attribution. Through extensive experiments involving open-source and proprietary large language models, we benchmark state-of-the-art document QA methods on VisDoMBench. Extensive results show that VisDoMRAG outperforms unimodal and long-context LLM baselines for end-to-end multimodal document QA by 12-20%.

Multimodal fusion learning has shown significant promise in classifying various diseases such as skin cancer and brain tumors. However, existing methods face three key limitations. First, they often lack generalizability to other diagnosis tasks due to their focus on a particular disease. Second, they do not fully leverage multiple health records from diverse modalities to learn robust complementary information. And finally, they typically rely on a single attention mechanism, missing the benefits of multiple attention strategies within and across various modalities. To address these issues, this paper proposes a dual robust information fusion attention mechanism (DRIFA) that leverages two attention modules, i.e. multi-branch fusion attention module and the multimodal information fusion attention module. DRIFA can be integrated with any deep neural network, forming a multimodal fusion learning framework denoted as DRIFA-Net. We show that the multi-branch fusion attention of DRIFA learns enhanced representations for each modality, such as dermoscopy, pap smear, MRI, and CT-scan, whereas multimodal information fusion attention module learns more refined multimodal shared representations, improving the network's generalization across multiple tasks and enhancing overall performance. Additionally, to estimate the uncertainty of DRIFA-Net predictions, we have employed an ensemble Monte Carlo dropout strategy. Extensive experiments on five publicly available datasets with diverse modalities demonstrate that our approach consistently outperforms state-of-the-art methods. The code is available at https://github.com/misti1203/DRIFA-Net., Comment: 10 pages

In this work, we leverage the Hamiltonian kind structure for accurate uncertainty propagation through a nonlinear dynamical system. The developed approach utilizes the fact that the stationary probability density function is purely a function of the Hamiltonian of the system. This fact is exploited to define the basis functions for approximating the solution of the Fokker-Planck-Kolmogorov equation. This approach helps in curtailing the growth of basis functions with the state dimension. Furthermore, sparse approximation tools have been utilized to automatically select appropriate basis functions from an over-complete dictionary. A nonlinear oscillator and two-body problem are considered to show the efficacy of the proposed approach. Simulation results show that such an approach is effective in accurately propagating uncertainty through non-conservative as well as conservative systems., Comment: AAS 23-361

Existing LLM agent systems typically select actions from a fixed and predefined set at every step. While this approach is effective in closed, narrowly-scoped environments, we argue that it presents two major challenges when deploying LLM agents in real-world scenarios: (1) selecting from a fixed set of actions significantly restricts the planning and acting capabilities of LLM agents, and (2) this approach requires substantial human effort to enumerate and implement all possible actions, which becomes impractical in complex environments with a vast number of potential actions. In this work, we propose an LLM agent framework that enables the dynamic creation and composition of actions in an online manner. In this framework, the agent interacts with the environment by generating and executing programs written in a general-purpose programming language at each step. Furthermore, generated actions are accumulated over time for future reuse. Our extensive experiments on the GAIA benchmark demonstrate that this framework offers significantly greater flexibility and outperforms previous methods. Notably, it allows an LLM agent to recover in scenarios where no relevant action exists in the predefined set or when existing actions fail due to unforeseen edge cases. At the time of writing, we hold the top position on the GAIA public leaderboard. Our code can be found in \href{https://github.com/adobe-research/dynasaur}{https://github.com/adobe-research/dynasaur}., Comment: 15 pages, 8 figures

The applications of generative AI have become extremely impressive, and the interplay between users and AI is even more so. Current human-AI interaction literature has taken a broad look at how humans interact with generative AI, but it lacks specificity regarding the user interface designs and patterns used to create these applications. Therefore, we present a survey that comprehensively presents taxonomies of how a human interacts with AI and the user interaction patterns designed to meet the needs of a variety of relevant use cases. We focus primarily on user-guided interactions, surveying interactions that are initiated by the user and do not include any implicit signals given by the user. With this survey, we aim to create a compendium of different user-interaction patterns that can be used as a reference for designers and developers alike. In doing so, we also strive to lower the entry barrier for those attempting to learn more about the design of generative AI applications.

Efficient long-context language modeling remains a significant challenge in Natural Language Processing (NLP). While Transformers dominate language tasks, they struggle with long sequences due to quadratic computational complexity in training and linearly scaling memory costs during inference. Recent State Space Models (SSMs) such as Mamba offer alternatives with constant memory usage, but they underperform in tasks requiring extensive in-context retrieval. We introduce Taipan, a novel hybrid architecture that combines Mamba-2 with Selective Attention Layers (SALs). These SALs identify tokens requiring long-range interactions, remove less important features, and then augment their representations using the attention module. This approach balances Mamba's efficiency with Transformer-like performance in memory-intensive tasks. By constraining the attention budget, Taipan extends accurate predictions to context lengths of up to 1 million tokens while preserving computational efficiency. Our experiments demonstrate Taipan's superior performance across various scales and tasks, offering a promising solution for efficient long-context language modeling.

Document structure editing involves manipulating localized textual, visual, and layout components in document images based on the user's requests. Past works have shown that multimodal grounding of user requests in the document image and identifying the accurate structural components and their associated attributes remain key challenges for this task. To address these, we introduce the DocEdit-v2, a novel framework that performs end-to-end document editing by leveraging Large Multimodal Models (LMMs). It consists of three novel components: (1) Doc2Command, which simultaneously localizes edit regions of interest (RoI) and disambiguates user edit requests into edit commands; (2) LLM-based Command Reformulation prompting to tailor edit commands originally intended for specialized software into edit instructions suitable for generalist LMMs. (3) Moreover, DocEdit-v2 processes these outputs via Large Multimodal Models like GPT-4V and Gemini, to parse the document layout, execute edits on grounded Region of Interest (RoI), and generate the edited document image. Extensive experiments on the DocEdit dataset show that DocEdit-v2 significantly outperforms strong baselines on edit command generation (2-33%), RoI bounding box detection (12-31%), and overall document editing (1-12\%) tasks., Comment: EMNLP 2024 (Main)

Glassy dynamics in a dense system of active particles with self-propulsion force $f_0$ and persistence time $\tau_p$ are crucial for many biological processes. Recent studies have shown that, unlike relaxation dynamics, dynamic heterogeneity (DH) in active glasses exhibits nontrivial behavior. However, the mechanism by which activity affects DH remains unknown. We have developed an active inhomogeneous mode-coupling theory (aIMCT) for DH in active glasses. We show that the nontrivial behavior of DH comes from a novel nonequilibrium effect of activity that leads to distinct behaviors of DH and relaxation dynamics in active glasses. When activity is small, DH exhibits equilibrium-like behavior with a power-law divergence of the peak height of the four-point correlation function, $\chi_C^\text{peak}$, and the aIMCT value of the exponent, $\mu\simeq 1.0$, is consistent with the existing and our new simulations of active glasses. However, $\chi_C^\text{peak}$ deviates from the scaling relations at higher $f_0$ values because of the novel effect on DH, although the deviation with varying $\tau_p$ is relatively weak.

In the high-stakes domain of search-and-rescue missions, the deployment of Unmanned Aerial Vehicles (UAVs) has become increasingly pivotal. These missions require seamless, real-time communication among diverse roles within response teams, particularly between Remote Operators (ROs) and On-Site Operators (OSOs). Traditionally, ROs and OSOs have relied on radio communication to exchange critical information, such as the geolocation of victims, hazardous areas, and points of interest. However, radio communication lacks information visualization, suffers from noise, and requires mental effort to interpret information, leading to miscommunications and misunderstandings. To address these challenges, this paper presents VizCom-AR, an Augmented Reality system designed to facilitate visual communication between ROs and OSOs and their situational awareness during UAV-driven search-and-rescue missions. Our experiments, focus group sessions with police officers, and field study showed that VizCom-AR enhances spatial awareness of both ROs and OSOs, facilitate geolocation information exchange, and effectively complement existing communication tools in UAV-driven emergency response missions. Overall, VizCom-AR offers a fundamental framework for designing Augmented Reality systems for large scale UAV-driven rescue missions., Comment: 8 pages

This study is the first to analyze the performance of a time-series foundation model for Value-at-Risk (VaR), which essentially forecasts the left-tail quantiles of returns. Foundation models, pre-trained on diverse datasets, can be applied in a zero-shot setting with minimal data or further improved through finetuning. We compare Google's TimesFM model to conventional parametric and non-parametric models, including GARCH and Generalized Autoregressive Score (GAS), using 19 years of daily returns from the S&P 100 index and its constituents. Backtesting with over 8.5 years of out-of-sample data shows that the fine-tuned foundation model consistently outperforms traditional methods in actual-over-expected ratios. For the quantile score loss function, it performs comparably to the best econometric model, GAS. Overall, the foundation model ranks as the best or among the top performers across the 0.01, 0.025, 0.05, and 0.1 quantile forecasting. Fine-tuning significantly improves accuracy, showing that zero-shot use is not optimal for VaR.

In this article, we employ physics-informed residual learning (PIRL) and propose a pricing method for European options under a regime-switching framework, where closed-form solutions are not available. We demonstrate that the proposed approach serves an efficient alternative to competing pricing techniques for regime-switching models in the literature. Specifically, we demonstrate that PIRLs eliminate the need for retraining and become nearly instantaneous once trained, thus, offering an efficient and flexible tool for pricing options across a broad range of specifications and parameters.

Quantitative evaluation of echocardiography is essential for precise assessment of cardiac condition, monitoring disease progression, and guiding treatment decisions. The diverse nature of echo images, including variations in probe types, manufacturers, and pathologies, poses challenges for developing artificial intelligent models that can generalize across different clinical practice. We introduce EchoApex, the first general-purpose vision foundation model echocardiography with applications on a variety of clinical practice. Leveraging self-supervised learning, EchoApex is pretrained on over 20 million echo images from 11 clinical centres. By incorporating task-specific decoders and adapter modules, we demonstrate the effectiveness of EchoApex on 4 different kind of clinical applications with 28 sub-tasks, including view classification, interactive structure segmentation, left ventricle hypertrophy detection and automated ejection fraction estimation from view sequences. Compared to state-of-the-art task-specific models, EchoApex attains improved performance with a unified image encoding architecture, demonstrating the benefits of model pretraining at scale with in-domain data. Furthermore, EchoApex illustrates the potential for developing a general-purpose vision foundation model tailored specifically for echocardiography, capable of addressing a diverse range of clinical applications with high efficiency and efficacy.

Cardiac magnetic resonance imaging (CMR), considered the gold standard for noninvasive cardiac assessment, is a diverse and complex modality requiring a wide variety of image processing tasks for comprehensive assessment of cardiac morphology and function. Advances in deep learning have enabled the development of state-of-the-art (SoTA) models for these tasks. However, model training is challenging due to data and label scarcity, especially in the less common imaging sequences. Moreover, each model is often trained for a specific task, with no connection between related tasks. In this work, we introduce a vision foundation model trained for CMR assessment, that is trained in a self-supervised fashion on 36 million CMR images. We then finetune the model in supervised way for 9 clinical tasks typical to a CMR workflow, across classification, segmentation, landmark localization, and pathology detection. We demonstrate improved accuracy and robustness across all tasks, over a range of available labeled dataset sizes. We also demonstrate improved few-shot learning with fewer labeled samples, a common challenge in medical image analyses. We achieve an out-of-box performance comparable to SoTA for most clinical tasks. The proposed method thus presents a resource-efficient, unified framework for CMR assessment, with the potential to accelerate the development of deep learning-based solutions for image analysis tasks, even with few annotated data available., Comment: 11 pages, 3 figures, 4 tables

Intra-cardiac echocardiography (ICE) is a crucial imaging modality used in electrophysiology (EP) and structural heart disease (SHD) interventions, providing realtime, high-resolution views from within the heart. Despite its advantages, effective manipulation of the ICE catheter requires significant expertise, which can lead to inconsistent outcomes, especially among less experienced operators. To address this challenge, we propose an AIdriven view guidance system that operates in a continuous closed-loop with human-in-the-loop feedback, designed to assist users in navigating ICE imaging without requiring specialized knowledge. Specifically, our method models the relative position and orientation vectors between arbitrary views and clinically defined ICE views in a spatial coordinate system. It guides users on how to manipulate the ICE catheter to transition from the current view to the desired view over time. By operating in a closedloop configuration, the system continuously predicts and updates the necessary catheter manipulations, ensuring seamless integration into existing clinical workflows. The effectiveness of the proposed system is demonstrated through a simulation-based performance evaluation using real clinical data, achieving an 89% success rate with 6,532 test cases. Additionally, a semi-simulation experiment with human-in-the-loop testing validated the feasibility of continuous yet discrete guidance. These results underscore the potential of the proposed method to enhance the accuracy and efficiency of ICE imaging procedures.

This paper presents VisioPhysioENet, a novel multimodal system that leverages visual cues and physiological signals to detect learner engagement. It employs a two-level approach for visual feature extraction using the Dlib library for facial landmark extraction and the OpenCV library for further estimations. This is complemented by extracting physiological signals using the plane-orthogonal-to-skin method to assess cardiovascular activity. These features are integrated using advanced machine learning classifiers, enhancing the detection of various engagement levels. We rigorously evaluate VisioPhysioENet on the DAiSEE dataset, where it achieves an accuracy of 63.09%, demonstrating a superior ability to discern various levels of engagement compared to existing methodologies. The proposed system's code can be accessed at https://github.com/MIntelligence-Group/VisioPhysioENet., Comment: 5 Pages, 2 figures

Time-modulated nanostructures allow us to control the spatial and temporal properties of light. The temporal modulation of the nanostructures constitutes an additional degree of freedom to control their scattering properties on demand and in a reconfigurable manner. However, these additional parameters create a vast design space, raising challenges in identifying optimal designs. Therefore, tools from the field of photonic inverse design must be used to optimize the degrees of freedom of the system to facilitate predefined optical responses. To further develop this field, here we introduce a differentiable transition (T-) matrix-based inverse design framework for dispersive time-modulated nanostructures. The electron density of the material of the nanostructures is modulated non-adiabatically as a generic periodic function of time. Using the inverse design framework, the temporal shape of the electron density can be manipulated to reach the target functionality. Our computational framework is exploited, exemplarily, in two instances. First, the decay rate enhancement of oscillating dipoles near time-modulated spheres is controlled on demand. Second, using spatiotemporal metasurfaces, a system supporting asymmetric transmission of light at visible frequencies is designed. Our work paves the way toward programmable spatiotemporal metasurfaces and space-time crystals for a future generation of reconfigurable functional photonic devices.

The transition matrix, frequently abbreviated as T-matrix, contains the complete information in a linear approximation of how a spatially localized object scatters an incident field. The T-matrix is used to study the scattering response of an isolated object and describes the optical response of complex photonic materials made from ensembles of individual objects. T-matrices of certain common structures, potentially, have been repeatedly calculated all over the world again and again. This is not necessary and constitutes a major challenge for various reasons. First, the resources spent on their computation represent an unsustainable financial and ecological burden. Second, with the onset of machine learning, data is the gold of our era, and it should be freely available to everybody to address novel scientific challenges. Finally, the possibility of reproducing simulations could tremendously improve if the considered T-matrices could be shared. To address these challenges, we found it important to agree on a common data format for T-matrices and to enable their collection from different sources and distribution. This document aims to develop the specifications for storing T-matrices and associated metadata. The specifications should allow maximum freedom to accommodate as many use cases as possible without introducing any ambiguity in the stored data. The common format will assist in setting up a public database of T-matrices., Comment: Submitted to the Journal of Quantitative Spectroscopy and Radiative Transfer

A hub-based colony consists of multiple agents who share a common nest site called the hub. Agents perform tasks away from the hub like foraging for food or gathering information about future nest sites. Modeling hub-based colonies is challenging because the size of the collective state space grows rapidly as the number of agents grows. This paper presents a graph-based representation of the colony that can be combined with graph-based encoders to create low-dimensional representations of collective state that can scale to many agents for a best-of-N colony problem. We demonstrate how the information in the low-dimensional embedding can be used with two experiments. First, we show how the information in the tensor can be used to cluster collective states by the probability of choosing the best site for a very small problem. Second, we show how structured collective trajectories emerge when a graph encoder is used to learn the low-dimensional embedding, and these trajectories have information that can be used to predict swarm performance.

Aligning future system design with the ever-increasing compute needs of large language models (LLMs) is undoubtedly an important problem in today's world. Here, we propose a general performance modeling methodology and workload analysis of distributed LLM training and inference through an analytical framework that accurately considers compute, memory sub-system, network, and various parallelization strategies (model parallel, data parallel, pipeline parallel, and sequence parallel). We validate our performance predictions with published data from literature and relevant industry vendors (e.g., NVIDIA). For distributed training, we investigate the memory footprint of LLMs for different activation re-computation methods, dissect the key factors behind the massive performance gain from A100 to B200 ($\sim$ 35x speed-up closely following NVIDIA's scaling trend), and further run a design space exploration at different technology nodes (12 nm to 1 nm) to study the impact of logic, memory, and network scaling on the performance. For inference, we analyze the compute versus memory boundedness of different operations at a matrix-multiply level for different GPU systems and further explore the impact of DRAM memory technology scaling on inference latency. Utilizing our modeling framework, we reveal the evolution of performance bottlenecks for both LLM training and inference with technology scaling, thus, providing insights to design future systems for LLM training and inference., Comment: 12 pages, 9 figures

Safety fine-tuning helps align Large Language Models (LLMs) with human preferences for their safe deployment. To better understand the underlying factors that make models safe via safety fine-tuning, we design a synthetic data generation framework that captures salient aspects of an unsafe input by modeling the interaction between the task the model is asked to perform (e.g., "design") versus the specific concepts the task is asked to be performed upon (e.g., a "cycle" vs. a "bomb"). Using this, we investigate three well-known safety fine-tuning methods -- supervised safety fine-tuning, direct preference optimization, and unlearning -- and provide significant evidence demonstrating that these methods minimally transform MLP weights to specifically align unsafe inputs into its weights' null space. This yields a clustering of inputs based on whether the model deems them safe or not. Correspondingly, when an adversarial input (e.g., a jailbreak) is provided, its activations are closer to safer samples, leading to the model processing such an input as if it were safe. We validate our findings, wherever possible, on real-world models -- specifically, Llama-2 7B and Llama-3 8B., Comment: Preprint

AI Infrastructure plays a key role in the speed and cost-competitiveness of developing and deploying advanced AI models. The current demand for powerful AI infrastructure for model training is driven by the emergence of generative AI and foundational models, where on occasion thousands of GPUs must cooperate on a single training job for the model to be trained in a reasonable time. Delivering efficient and high-performing AI training requires an end-to-end solution that combines hardware, software and holistic telemetry to cater for multiple types of AI workloads. In this report, we describe IBM's hybrid cloud infrastructure that powers our generative AI model development. This infrastructure includes (1) Vela: an AI-optimized supercomputing capability directly integrated into the IBM Cloud, delivering scalable, dynamic, multi-tenant and geographically distributed infrastructure for large-scale model training and other AI workflow steps and (2) Blue Vela: a large-scale, purpose-built, on-premises hosting environment that is optimized to support our largest and most ambitious AI model training tasks. Vela provides IBM with the dual benefit of high performance for internal use along with the flexibility to adapt to an evolving commercial landscape. Blue Vela provides us with the benefits of rapid development of our largest and most ambitious models, as well as future-proofing against the evolving model landscape in the industry. Taken together, they provide IBM with the ability to rapidly innovate in the development of both AI models and commercial offerings., Comment: Corresponding Authors: Talia Gershon, Seetharami Seelam,Brian Belgodere, Milton Bonilla

Segmentation is often the first step in many medical image analyses workflows. Deep learning approaches, while giving state-of-the-art accuracies, are data intensive and do not scale well to low data regimes. We introduce Deep Conditional Shape Models 2.0, which uses an edge detector, along with an implicit shape function conditioned on edge maps, to leverage cross-modality shape information. The shape function is trained exclusively on a source domain (contrasted CT) and applied to the target domain of interest (3D echocardiography). We demonstrate data efficiency in the target domain by varying the amounts of training data used in the edge detection stage. We observe that DCSM 2.0 outperforms the baseline at all data levels in terms of Hausdorff distances, and while using 50% or less of the training data in terms of average mesh distance, and at 10% or less of the data with the dice coefficient. The method scales well to low data regimes, with gains of up to 5% in dice coefficient, 2.58 mm in average surface distance and 21.02 mm in Hausdorff distance when using just 2% (22 volumes) of the training data., Comment: Best oral paper award at ISBI 2024

Distributed Deep Neural Network (DNN) training is a technique to reduce the training overhead by distributing the training tasks into multiple accelerators, according to a parallelization strategy. However, high-performance compute and interconnects are needed for maximum speed-up and linear scaling of the system. Wafer-scale systems are a promising technology that allows for tightly integrating high-end accelerators with high-speed wafer-scale interconnects, making it an attractive platform for distributed training. However, the wafer-scale interconnect should offer high performance and flexibility for various parallelization strategies to enable maximum optimizations for compute and memory usage. In this paper, we propose FRED, a wafer-scale interconnect that is tailored for the high-BW requirements of wafer-scale networks and can efficiently execute communication patterns of different parallelization strategies. Furthermore, FRED supports in-switch collective communication execution that reduces the network traffic by approximately 2X. Our results show that FRED can improve the average end-to-end training time of ResNet-152, Transformer-17B, GPT-3, and Transformer-1T by 1.76X, 1.87X, 1.34X, and 1.4X, respectively when compared to a baseline waferscale 2D-Mesh fabric.

While the generation of document layouts has been extensively explored, comprehensive document generation encompassing both layout and content presents a more complex challenge. This paper delves into this advanced domain, proposing a novel approach called DocSynthv2 through the development of a simple yet effective autoregressive structured model. Our model, distinct in its integration of both layout and textual cues, marks a step beyond existing layout-generation approaches. By focusing on the relationship between the structural elements and the textual content within documents, we aim to generate cohesive and contextually relevant documents without any reliance on visual components. Through experimental studies on our curated benchmark for the new task, we demonstrate the ability of our model combining layout and textual information in enhancing the generation quality and relevance of documents, opening new pathways for research in document creation and automated design. Our findings emphasize the effectiveness of autoregressive models in handling complex document generation tasks., Comment: Spotlight (Oral) Acceptance to CVPR 2024 Workshop for Graphic Design Understanding and Generation (GDUG)