Enterprise-grade Intelligent Document Processing (IDP) systems support high-stakes workflows across finance, insurance, and healthcare. Early-phase system validation under limited budgets mandates uncovering diverse failure mechanisms, rather than identifying a single worst-case document. We formalize this challenge as a Search-Based Software Testing (SBST) problem, aiming to identify complex interactions between document variables, with the objective to maximize the number of distinct failure types discovered within a fixed evaluation budget. Our methodology operates on a combinatorial space of document configurations, rendering instances of structural risk features to induce realistic failure conditions. We benchmark a diverse portfolio of search strategies spanning evolutionary, swarm-based, quality-diversity, learning-based, and quantum under identical budget constraints. Through configuration-level exclusivity, win-rate, and cross-temporal overlap analyses, we show that different solvers consistently uncover failure modes that remain undiscovered by specific alternatives at comparable budgets. Crucially, cross-temporal analysis reveals persistent solver-specific discoveries across all evaluated budgets, with no single strategy exhibiting absolute dominance. While the union of all solvers eventually recovers the observed failure space, reliance on any individual method systematically delays the discovery of important risks. These results demonstrate intrinsic solver complementarity and motivate portfolio-based SBST strategies for robust industrial IDP validation.

Lead optimization is a pivotal task in the drug design phase within the drug discovery lifecycle. The primary objective is to refine the lead compound to meet specific molecular properties for progression to the subsequent phase of development. In this work, we present an innovative approach, Latent Evolutionary Optimization for Molecule Generation (LEOMol), a generative modeling framework for the efficient generation of op-timized molecules. LEO Mol leverages Evolutionary Algorithms, such as Genetic Algorithm and Differential Evolution, to search the latent space of a Variational AutoEncoder (VAE). This search facilitates the identification of the target molecule distribution within the latent space. Our approach consistently demonstrates superior performance compared to previous state-of-the-art mod-els across a range of constrained molecule generation tasks, outperforming existing models in all four sub-tasks related to property targeting. Additionally, we suggest the importance of including toxicity in the evaluation of generative models. Furthermore, an ablation study underscores the improvements that our approach provides over gradient-based latent space optimization methods. This underscores the effectiveness and superiority of LEO Mol in addressing the inherent challenges in constrained molecule generation while emphasizing its potential to propel advancements in drug discovery.

Large Language Models (LLMs) have demonstrated remarkable capabilities in various language-related tasks enabling applications in various fields such as healthcare, education, financial services etc. However, they are prone to producing factually incorrect responses or ''hallucinations'' which can have detrimental consequences such as loss of credibility, diminished customer trust etc. In this presentation, we showcase a solution that addresses the challenge of minimizing hallucinations. Our solution provides accurate responses and generates detailed explanations, thereby enabling the users to know how the model arrived at the final response. Additionally, it verifies if the explanations are factually correct and offers insights into whether the generated explanations are directly derived from the provided context or if they are inferred from it. We also systematically assess the quality of generated responses using an LLM-based evaluation technique. We present empirical results on benchmark datasets to demonstrate the effectiveness of our approach. Our presentation also examines the impact of individual components in the solution, enhancing the factual correctness of the final response. This research is vital for industries utilizing LLMs, as it provides a means to enhance the reliability of responses and mitigate the risks associated with factual hallucinations. Researchers and practitioners seeking to enhance the reliability of LLM responses will find valuable insights in this presentation.

Pharmacovigilance is the science and practice of monitoring, assessing, and preventing adverse effects or any other drug-related problems. Pharmacovigilance ensures the post-market safety of pharmaceuticals and plays a crucial role in public health by enhancing drug safety. This discipline involves collecting, analyzing, and reporting data on adverse events, allowing for informed regulatory decisions. Manual systems face challenges in handling data volume, potentially leading to oversight and delays. Automation with advanced technologies can be a practical solution to mitigate these challenges and ensure efficient data management.

Foundational Large Language Models (LLMs) are pre-trained generally on huge corpora encompassing broad subjects to become versatile and generalize to future downstream tasks. However, their effectiveness falls short when dealing with tasks that are highly specialized to a specific use case. Even when adopting current prompt engineering techniques like few-shot or Chain-of-Thought reasoning prompts, the required level of results is not yet achievable directly with foundational models alone. The alternative approach is to fine-tune the LLM, but a common challenge is the limited availability of task-specific training data. In this talk, we will introduce an end-to-end automated framework to tailor a model to specific downstream tasks for an industry where the first step is to generate task-specific custom data from unstructured documents. Next, we will discuss our optimized distributed training pipeline for fine-tuning LLMs on the generated data. Finally, we will provide an overview of the statistical metrics and customized metrics we employ for assessing the performance of the fine-tuned LLM. This automated framework alleviates the burden of manual adjustments and streamlines the process to provide a model that is fully customized to suit the unique requirements of any specific business use case.

Deep neural networks model data for a task or a sequence of tasks, where the knowledge extracted from the data is encoded in the parameters and representations of the network. Extraction and utilization of these representations is vital when data is no longer available in the future, especially in a continual learning scenario. We introduce flashcards, which are visual representations that capture the encoded knowledge of a network as a recursive function of some predefined random image patterns. In a continual learning scenario, flashcards help to prevent catastrophic forgetting by consolidating the knowledge of all the previous tasks. Flashcards are required to be constructed only before learning the subsequent task, hence, they are independent of the number of tasks trained before, making them task agnostic. We demonstrate the efficacy of flashcards in capturing learned knowledge representation (as an alternative to the original data), and empirically validate on a variety of continual learning tasks: reconstruction, denoising, and task-incremental classification, using several heterogeneous (varying background and complexity) benchmark datasets. Experimental evidence indicates that: (i) flashcards as a replay strategy is task agnostic, (ii) performs better than generative replay, and (iii) is on par with episodic replay without additional memory overhead.

Continual learning (CL) aims to learn new tasks by forward transfer of information learnt from previous tasks and without forgetting them. In task incremental CL, task information is vital during both strategy development and inference. Providing such partial knowledge about the test sample demands additional complexity and may become intractable, especially when the sample source is ambiguous. In this work, we design a task-agnostic approach that uses base-child hybrid setup to incrementally learn tasks while mitigating forgetting. Multiple base classifiers guided by reference points learn new tasks and this information is distilled via feature space induced sampling strategy. A central child classifier consolidates information across tasks and infers the task identifier automatically. Experimental results on standard datasets show that the proposed approach outperforms the various state-of-the-art regularization and replay CL algorithms in terms of accuracy, by 50% and 7% with homogeneous and heterogeneous tasks, respectively, in task-agnostic scenarios.

This paper demonstrates the use of 3D Anisotropic Convolutional Neural Network (CNN) with predict-refine mechanism for 3D brain tumor segmentation. We propose two networks that utilize multi-scale feedback and saliency maps respectively to segment three critical regions involved in automated brain tumor segmentation. The proposed networks are formulated to predict feature maps at different resolutions during the prediction phase. These networks perform refinement process using the saliency or feature maps as feedback information for the refinement process. The recurrent architecture allows the network to automatically rectify errors in saliency map of the previous prediction phase resulting in more reliable final predictions. Our experimental results on the BraTS2017 dataset demonstrate the superior performance of our proposed predict-refine architecture than current state of the art approaches improving results by up to 8% without any additional increase in the 1.9M model parameters.

Quantitative measurements obtained from medical images guide clinicians in several use cases but manually obtaining such measurements are both laborious and subject to inter-observer variations. We develop a hybrid deep reinforced regression framework to robustly measure the Cardio-Thoracic ratio (CTR) from Chest X-ray (CXR) images, thereby directly identifying the presence of Cardiomegaly. The proposed hybrid framework initially employs a CNN based Regressor on pre-processed images to obtain approximate critical points. As the actual critical points are based on human expert's experience and subject to labeling uncertainties, a deep reinforcement learning (deep RL) approach is specifically designed to fine-tune estimated regression points from the CNN Regressor. The final regressed points are then used to measure CTR. Wingspan and ChestX-ray8 datasets are used for validating the proposed framework. The proposed framework shows generalization ability on ChestX-ray8 and outperforms the state-of-the-art results on Wingspan.