Shweta Singh

Shweta Singh, Aayan Yadav, Jitesh Jain et al. · gatech

The Common Objects in Context (COCO) dataset has been instrumental in benchmarking object detectors over the past decade. Like every dataset, COCO contains subtle errors and imperfections stemming from its annotation procedure. With the advent of high-performing models, we ask whether these errors of COCO are hindering its utility in reliably benchmarking further progress. In search for an answer, we inspect thousands of masks from COCO (2017 version) and uncover different types of errors such as imprecise mask boundaries, non-exhaustively annotated instances, and mislabeled masks. Due to the prevalence of COCO, we choose to correct these errors to maintain continuity with prior research. We develop COCO-ReM (Refined Masks), a cleaner set of annotations with visibly better mask quality than COCO-2017. We evaluate fifty object detectors and find that models that predict visually sharper masks score higher on COCO-ReM, affirming that they were being incorrectly penalized due to errors in COCO-2017. Moreover, our models trained using COCO-ReM converge faster and score higher than their larger variants trained using COCO-2017, highlighting the importance of data quality in improving object detectors. With these findings, we advocate using COCO-ReM for future object detection research. Our dataset is available at https://cocorem.xyz

Srihari R, Adarsha B, Mohammed Usman Hussain et al.

Users of government employment websites commonly face engagement and accessibility challenges linked to navigational complexity, a dearth of language options, and a lack of personalized support. This paper introduces JobSphere, an AI-powered career assistant that is redefining the employment platform in Punjab called PGRKAM. JobSphere employs Retrieval-Augmented Generation (RAG) architecture, and it is multilingual, available in English, Hindi and Punjabi. JobSphere technique uses 4-bit quantization, allowing the platform to deploy on consumer-grade GPUs (i.e., NVIDIA RTX 3050 4GB), making the implementation 89% cheaper than that of cloud-based systems. Key innovations include voice-enabled interaction with the assistant, automated mock tests, resume parsing with skills recognition, and embed-based job recommendation that achieves a precision@10 score of 68%. An evaluation of JobSphere's implementation reveals 94% factual accuracy, a median response time of 1.8 seconds, and a System Usability Scale score of 78.5/100, a 50% improvement compared to the baseline PGRKAM platform context. In conclusion, JobSphere effectively fills significant accessibility gaps for Punjab/Hindi-speaking users in rural locations, while also affirming the users access to trusted job content provided by government agencies.

Shivank Garg, Ayush Singh, Shweta Singh et al.

Reinforcement learning from human feedback (RLHF) has emerged as the primary method for aligning large language models (LLMs) with human preferences. While it enables LLMs to achieve human-level alignment, it often incurs significant computational and financial costs due to its reliance on training external reward models or human-labeled preferences. In this work, we propose Implicit Preference Optimization (IPO), an alternative approach that leverages generative LLMs as preference classifiers, thereby reducing the dependence on external human feedback or reward models to obtain preferences. We conduct a comprehensive evaluation on the preference classification ability of LLMs using RewardBench, assessing models across different sizes, architectures, and training levels to validate our hypothesis. Furthermore, we investigate the self-improvement capabilities of LLMs by generating multiple responses for a given instruction and employing the model itself as a preference classifier for Direct Preference Optimization (DPO)-based training. Our findings demonstrate that models trained through IPO achieve performance comparable to those utilizing state-of-the-art reward models for obtaining preferences.

Arti Rani, Shweta Singh, Nihar Ranjan Sahoo et al.

Large Language Models (LLMs) have achieved remarkable success on question answering (QA) tasks, yet they often encode harmful biases that compromise fairness and trustworthiness. Most existing bias mitigation approaches are restricted to predefined categories, limiting their ability to address novel or context-specific emergent biases. To bridge this gap, we tackle the novel problem of open-set bias detection and mitigation in text-based QA. We introduce OpenBiasBench, a comprehensive benchmark designed to evaluate biases across a wide range of categories and subgroups, encompassing both known and previously unseen biases. Additionally, we propose Open-DeBias, a novel, data-efficient, and parameter-efficient debiasing method that leverages adapter modules to mitigate existing social and stereotypical biases while generalizing to unseen ones. Compared to the state-of-the-art BMBI method, Open-DeBias improves QA accuracy on BBQ dataset by nearly $48\%$ on ambiguous subsets and $6\%$ on disambiguated ones, using adapters fine-tuned on just a small fraction of the training data. Remarkably, the same adapters, in a zero-shot transfer to Korean BBQ, achieve $84\%$ accuracy, demonstrating robust language-agnostic generalization. Through extensive evaluation, we also validate the effectiveness of Open-DeBias across a broad range of NLP tasks, including StereoSet and CrowS-Pairs, highlighting its robustness, multilingual strength, and suitability for general-purpose, open-domain bias mitigation. The project page is available at: https://sites.google.com/view/open-debias25

William Farlessyost, Sebastian Oberst, Shweta Singh

This paper introduces a novel framework for optimizing observer-based soft sensors through dynamic causality analysis. Traditional approaches to sensor selection often rely on linearized observability indices or statistical correlations that fail to capture the temporal evolution of complex systems. We address this gap by leveraging liquid-time constant (LTC) networks, continuous-time neural architectures with input-dependent time constants, to systematically identify and prune sensor inputs with minimal causal influence on state estimation. Our methodology implements an iterative workflow: training an LTC observer on candidate inputs, quantifying each input's causal impact through controlled perturbation analysis, removing inputs with negligible effect, and retraining until performance degradation occurs. We demonstrate this approach on three mechanistic testbeds representing distinct physical domains: a harmonically forced spring-mass-damper system, a nonlinear continuous stirred-tank reactor, and a predator-prey model following the structure of the Lotka-Volterra model, but with seasonal forcing and added complexity. Results show that our causality-guided pruning consistently identifies minimal sensor sets that align with underlying physics while improving prediction accuracy. The framework automatically distinguishes essential physical measurements from noise and determines when derived interaction terms provide complementary versus redundant information. Beyond computational efficiency, this approach enhances interpretability by grounding sensor selection decisions in dynamic causal relationships rather than static correlations, offering significant benefits for soft sensing applications across process engineering, ecological monitoring, and agricultural domains.

Santanu Roy, Shweta Singh, Palak Sahu et al.

Lung cancer, a severe form of malignant tumor that originates in the tissues of the lungs, can be fatal if not detected in its early stages. It ranks among the top causes of cancer-related mortality worldwide. Detecting lung cancer manually using chest X-Ray image or Computational Tomography (CT) scans image poses significant challenges for radiologists. Hence, there is a need for automatic diagnosis system of lung cancers from radiology images. With the recent emergence of deep learning, particularly through Convolutional Neural Networks (CNNs), the automated detection of lung cancer has become a much simpler task. Nevertheless, numerous researchers have addressed that the performance of conventional CNNs may be hindered due to class imbalance issue, which is prevalent in medical images. In this research work, we have proposed a novel CNN architecture ``Multi-Scale Dense Network (MSD-Net)'' (trained-from-scratch). The novelties we bring in the proposed model are (I) We introduce novel dense modules in the 4th block and 5th block of the CNN model. We have leveraged 3 depthwise separable convolutional (DWSC) layers, and one 1x1 convolutional layer in each dense module, in order to reduce complexity of the model considerably. (II) Additionally, we have incorporated one skip connection from 3rd block to 5th block and one parallel branch connection from 4th block to Global Average Pooling (GAP) layer. We have utilized dilated convolutional layer (with dilation rate=2) in the last parallel branch in order to extract multi-scale features. Extensive experiments reveal that our proposed model has outperformed latest CNN model ConvNext-Tiny, recent trend Vision Transformer (ViT), Pooling-based ViT (PiT), and other existing models by significant margins.

Khush Mendiratta, Shweta Singh, Pratik Chattopadhyay

Early detection of brain tumors through magnetic resonance imaging (MRI) is essential for timely treatment, yet access to diagnostic facilities remains limited in remote areas. Gliomas, the most common primary brain tumors, arise from the carcinogenesis of glial cells in the brain and spinal cord, with glioblastoma patients having a median survival time of less than 14 months. MRI serves as a non-invasive and effective method for tumor detection, but manual segmentation of brain MRI scans has traditionally been a labor-intensive task for neuroradiologists. Recent advancements in computer-aided design (CAD), machine learning (ML), and deep learning (DL) offer promising solutions for automating this process. This study proposes an automated deep learning model for brain tumor detection and classification using MRI data. The model, incorporating spatial attention, achieved 96.90% accuracy, enhancing the aggregation of contextual information for better pattern recognition. Experimental results demonstrate that the proposed approach outperforms baseline models, highlighting its robustness and potential for advancing automated MRI-based brain tumor analysis.

William Farlessyost, Shweta Singh

Dynamical analysis of manufacturing and natural systems provides critical information about production of manufactured and natural resources respectively, thus playing an important role in assessing sustainability of these systems. However, current dynamic models for these systems exist as mechanistic models, simulation of which is computationally intensive and does not provide a simplified understanding of the mechanisms driving the overall dynamics. For such systems, lower-order models can prove useful to enable sustainability analysis through coupled dynamical analysis. There have been few attempts at finding low-order models of manufacturing and natural systems, with existing work focused on model development of individual mechanism level. This work seeks to fill this current gap in the literature of developing simplified dynamical models for these systems by developing reduced-order models using a machine learning (ML) approach. The approach is demonstrated on an entire soybean-oil to soybean-diesel process plant and a lake system. We use a grey-box ML method with a standard nonlinear optimization approach to identify relevant models of governing dynamics as ODEs using the data simulated from mechanistic models. Results show that the method identifies a high accuracy linear ODE models for the process plant, reflective of underlying linear stoichiometric mechanisms and mass balance driving the dynamics. For the natural systems, we modify the ML approach to include the effect of past dynamics, which gives non-linear ODE. While the modified approach provides a better match to dynamics of stream flow, it falls short of completely recreating the dynamics. We conclude that the proposed ML approach work well for systems where dynamics is smooth, such as in manufacturing plant whereas does not work perfectly well in case of chaotic dynamics such as water stream flow.

Renganathan Subramanian, Shweta Singh

Machine learning recently has been used to identify the governing equations for dynamics in physical systems. The promising results from applications on systems such as fluid dynamics and chemical kinetics inspire further investigation of these methods on complex engineered systems. Dynamics of these systems play a crucial role in design and operations. Hence, it would be advantageous to learn about the mechanisms that may be driving the complex dynamics of systems. In this work, our research question was aimed at addressing this open question about applicability and usefulness of novel machine learning approach in identifying the governing dynamical equations for engineered systems. We focused on distillation column which is an ubiquitous unit operation in chemical engineering and demonstrates complex dynamics i.e. it's dynamics is a combination of heuristics and fundamental physical laws. We tested the method of Sparse Identification of Non-Linear Dynamics (SINDy) because of it's ability to produce white-box models with terms that can be used for physical interpretation of dynamics. Time series data for dynamics was generated from simulation of distillation column using ASPEN Dynamics. One promising result was reduction of number of equations for dynamic simulation from 1000s in ASPEN to only 13 - one for each state variable. Prediction accuracy was high on the test data from system within the perturbation range, however outside perturbation range equations did not perform well. In terms of physical law extraction, some terms were interpretable as related to Fick's law of diffusion (with concentration terms) and Henry's law (with ratio of concentration and pressure terms). While some terms were interpretable, we conclude that more research is needed on combining engineering systems with machine learning approach to improve understanding of governing laws for unknown dynamics.