Abstract

We develop new algorithms for Quantum Singular Value Transformation (QSVT), a unifying framework underlying a wide range of quantum algorithms. Existing implementations of QSVT rely on block encoding, incurring $O(log L)$ ancilla overhead and circuit depth $widetilde{O}(d lambda L)$ for polynomial transformations of a Hamiltonian $H = sum_{k=1}^L lambda_k H_k$, where $d$ is the polynomial degree, and $lambda = sum_k |lambda_k|$. We introduce a new approach that eliminates block encoding, needs only a single ancilla qubit, and maintains near-optimal complexity, using only basic Hamiltonian simulation methods such as Trotterization. Our method achieves a circuit depth of $widetilde{O}left(L (d lambda_{text{comm}})^{1+o(1)}right)$, without any multi-qubit controlled gates. Here, $lambda_{text{comm}}$ depends on the nested commutators of the $H_k$'s and can be much smaller than $lambda$. Central to our technique is a novel use of Richardson extrapolation, enabling systematic error cancellation in interleaved sequences of arbitrary unitaries and Hamiltonian evolution operators, establishing a broadly applicable framework beyond QSVT. Additionally, we propose two randomized QSVT algorithms for cases with only sampling access to Hamiltonian terms. The first uses qDRIFT, while the second replaces block encodings in QSVT with randomly sampled unitaries. Both achieve quadratic complexity in $d$, which we establish as a lower bound for any randomized method implementing polynomial transformations in this model. Finally, as applications, we develop end-to-end quantum algorithms for quantum linear systems and ground state property estimation, achieving near-optimal complexity without oracular access. Our results provide a new framework for quantum algorithms, reducing hardware overhead while maintaining near-optimal performance, with implications for both near-term and fault-tolerant quantum computing.

Abstract

Recent evaluations of LLMs on coreference resolution have revealed that traditional output formats and evaluation metrics do not fully capture the models' referential understanding. To address this, we introduce IdentifyMe, a new benchmark for mention resolution presented in a multiple-choice question (MCQ) format, commonly used for evaluating LLMs. IdentifyMe features long narratives and employs heuristics to exclude easily identifiable mentions, creating a more challenging task. The benchmark also consists of a curated mixture of different mention types and corresponding entities, allowing for a fine-grained analysis of model performance. We evaluate both closed- and open source LLMs on IdentifyMe and observe a significant performance gap (20-30%) between the state-of-the-art sub-10B open models vs. closed ones. We observe that pronominal mentions, which have limited surface information, are typically much harder for models to resolve than nominal mentions. Additionally, we find that LLMs often confuse entities when their mentions overlap in nested structures. The highest-scoring model, GPT-4o, achieves 81.9% accuracy, highlighting the strong referential capabilities of state-of-the-art LLMs while also indicating room for further improvement

Abstract

Parkinson’s Disease (PD) neurodegenerative disorder which lacks reliable early diagnostic tests. In this study, we present a method for PD prediction that uses multimodal data from the Parkinson Progression Marker Initiative (PPMI), specifically clinico-demographic, biospecimen, and genetic data, to improve predictive accuracy and facilitate timely interventions via a multimodal machine learning approach. We evaluated data obtained from 598 participants (171 healthy controls and 427 PD patients) in three different modalities: 29 clinical features, five cerebrospinal fluid biomarkers, and 154 SNPs (single nucleotide polymorphisms), which were selected through a biology-driven feature selection method. We utilized three multimodal integration strategies—early, intermediate, and late—and trained various machine learning models, including LightGBM and Multilayer Perceptron (MLP), each of which was optimized by hyperparameter tuning and cross-validation. In early integration, we combined feature sets from all modalities into a single set, leveraging complementary information to increase predictive power. The intermediate integration method made use of autoencoders to encode features into a single 12 dimensional vector as the input into a Neural Network classifier. Late integration combined outputs from the top-performing models for each modality using ensemble techniques such as Voting Classifier and Stacking. We found that early integration achieved the highest performance, with Support Vector Machines with 90% accuracy, 0.98 AUC-ROC, 0.99 precision, and 0.93 F1 score. Intermediate integration with the Neural Network classifier followed with an AUC-ROC of 0.84 and an F1-score of 0.85. Our feature importance analysis identified two clinical scores, namely UPSIT (related to sensory decline) and SCOPA (measuring autonomic dysfunction), along with two SNPs (chr4 90755939 A G and chr4 90646886 G A, both previously linked to PD susceptibility) as crucial predictors, emphasizing their well established relevance in PD diagnosis. Early diagnosis and treatment of PD patients are facilitated by the integration of multiple data modalities, which also greatly increases the predicted accuracy for the disease while also providing us with a thorough understanding of its complexity.

Abstract

The constant shifts in social and political contexts, driven by emerging social movements and political events, lead to new forms of hate content and previously unrecognized hate patterns that machine learning models may not have captured. Some recent literature proposes data augmentation-based techniques to enrich existing hate datasets by incorporating samples that reveal new implicit hate patterns. This approach aims to improve the model's performance on out-of-domain implicit hate instances. It is observed, that further addition of more samples for augmentation results in the decrease of the performance of the model. In this work, we propose a Knowledge Transfer-driven Concept Refinement method that distills and refines the concepts related to implicit hate samples through novel prototype alignment and concept losses, alongside data augmentation based on concept activation vectors. Experiments with several publicly available datasets show that incorporating additional implicit samples reflecting new hate patterns through concept refinement enhances the model's performance, surpassing baseline results while maintaining cross-dataset generalization capabilities.

Abstract

In this work, we present IDD-CRS, a large-scale dataset focused on critical road scenarios, captured using Advanced Driver Assistance Systems (ADAS) and dash cameras. Unlike existing datasets that predominantly emphasize pedestrian safety and vehicle safety separately, IDD-CRS incorporates both vehicle and pedestrian behaviors, offering a more comprehensive view of road safety. The dataset includes diverse scenarios, such as high-speed lane changes, unsafe vehicle approaches to pedestrians and cyclists, and complex interactions between ego vehicles and other road agents. Leveraging ADAS technology allows us to accurately define the temporal boundaries of actions, resulting in precise annotations and more reliable safety analysis. With 90 hours of video footage, consisting of 5400 one-minute-long videos and 135,000 frames, IDD-CRS introduces new vehicle related classes and hard negative classes, establishing baselines for action recognition and long-tail action recognition tasks. Our benchmarks reveal the limitations of current models, pointing toward future advancements needed for improving road safety technology.

Abstract

This study uses a transcriptomic and machine learning-based approach to gain a comprehensive understanding of the genomic landscape of Plasmodium falciparum, a malaria-causing parasite. This enables us to understand this life-threatening disease better and thus identify new treatment targets and intervention methods. We used single-cell RNA sequencing data from the Malarial Cell Atlas with 5,177 genes across 37,624 cells. Our approach combined static analyses, including highly variable gene identification, differential expression analysis, and M3Drop (a feature selection method based on noise filtration), to obtain stage-specific markers along with trajectory analysis (Monocle3) to identify top genes active during stage progression. The feature selection analyses were done for each stage and lifecycle gene, which are genes that remain active throughout the lifecycle. These complementary approaches gave us distinct gene sets: static analysis revealed genes necessary for stage-specific functions, while dynamic trajectory analysis highlighted genes that play an important role in cellular development. A neural network classifier trained on these gene sets achieved high accuracy (Accuracy: 96%; F1-score: 0.96) using trajectory-derived genes, performing better than the model trained on the entire gene set, indicating that the gene set has captured stage-specific signatures. Further, to validate the biological significance of the shortlisted genes, we performed a Gene Ontology analysis. We found that our analysis revealed results that are in line with existing literature. The ring stage genes were linked to immune evasion, trophozoite genes to haemoglobin digestion, schizont genes to merozoite invasion, and gametocyte genes to sexual differentiation. We plan to build on this work by studying the behaviour of important genes using a Poisson-Beta model to infer the kinetics of key genes, which will help us better understand how they change throughout the parasite's life cycle.

Abstract

The rising cost of drug discovery, coupled with a stagnation in the approval of novel treatments, highlights the urgent need for innovative strategies such as drug re-purposing. Pharmaceutical companies invest roughly 10-15 years and $2.6 billion to get a single FDA-approved drug to market. The COVID-19 pandemic further underscored the necessity of quickly identifying existing drugs with potential efficacy against a fast-spreading virus to curtail the pandemic. In this study, we perform a comparative analysis of several Graph Neural Networks (GNNs) and recommendation system models to address drug re-purposing. We construct an integrated graph that combines Protein-Protein Interaction networks, Drug-Target Protein graphs, Disease-Protein associations, and Drug-Disease links. We leverage a network learning paradigm implemented over this complex graph via both node-agnostic and heterogeneous graph techniques for link prediction in drug-disease pairs. We implement a Heterogeneous Graph Transformer (HGT) model that processes three node types (drugs, diseases, proteins) and four edge types. The HGT achieved an AUC-ROC of 0.985 and an F1-score of 0.90, demonstrating its efficacy in predicting drug re-purposing candidates. Additionally, we compared several node-agnostic GNN architectures, including Graph Convolutional Networks, Graph Attention Networks, GraphSAGE, and Graph Isomorphism Networks. All architectures performed comparably, with an AUC-ROC of around 0.98. However, when framing the drug re-purposing task as a recommendation problem using Matrix Factorization with side information, we observed a significant drop in performance, with the AUC-ROC falling to 0.82. This performance degradation highlights the importance of incorporating Protein-Protein Interaction networks in the modelling process, as matrix factorization fails to capture these complex network effects critical for drug re-purposing. Our models ranked 6,158 drugs based on their predicted efficacy in treating COVID-19, providing a valuable tool for prioritizing clinical trials and further research. Beyond COVID-19, such an integrated framework can allow us to uncover drug re-purposing prospects for any other novel diseases in a significantly more efficient and cost-effective way.

Abstract

Gliomas, accounting for nearly one-third of all brain tumours, present themselves as significant medical challenges owing to their heterogeneous and aggressive nature. Given that traditional diagnostic techniques often fall short of capturing the entire genetic landscape, there are minimal options currently present in personalised medicine, thus adding to existing challenges. In order to aid precision oncology, we present a machine learning (ML) framework for glioma subtype classification and mutation prediction using RNA sequencing data. We used a two-step feature selection procedure that included XGBoost and LASSO, leveraging RNA sequencing and somatic mutation data of 667 glioma samples from the Cancer Genome Atlas (TCGA) Pan-Cancer project to identify key genes that are associated with glioma subtypes and important mutations. Metrics such as AUROC, F1-Score, and Matthews Correlation Coefficient (MCC) were used to train and evaluate the supervised machine learning models, including Random Forest and Stochastic Gradient Descent. Our framework performed robustly despite the class imbalance present in the dataset. We achieved a high classification accuracy for Diffuse Glioma (DG) subtypes (AUROC 0.90-1.00), distinguishing between low-grade gliomas (LGGs) and glioblastoma multiforme (GBM). Mutation status predictions for the key prognostic genes, including IDH, TP53, and ATRX, achieved high AUROC scores of 0.89-0.98 and MCC values of 0.72-0.97. Further, we used feature sets from our top-performing models to perform gene set enrichment as part of our post-analysis. This confirmed the biological significance of identified genes related to established carcinogenic pathways and further validated the clinical applicability of our technique. Our ML-driven framework presented offers a scalable, data-driven solution in computational oncology with potential applications across cancer types.

Abstract

Leptoquarks (LQs) can contribute to the magnetic and electric dipole moments of charged leptons, which the current experiments have measured with good accuracy. We revisit the parameter spaces of TeV-scale vector LQs that contribute to these observables and study how these models fare against the LHC bounds. We show that significant portions of the parameter space are excluded when the current LHC data is utilized effectively. We find that only U1and V2 can explain the observed positive shift in (aμexp-aμSM) through a lepton chirality-flipping contribution with O(1) LQ-quark-lepton coupling. We also see how these two LQs can fit the electron dipole moment and atomic parity violation measurements. We find that the current electric dipole moment measurements of the muon cannot restrain the LQ-quark-lepton couplings within perturbative regions.

Abstract

Despite the progress seen in classification methods, current approaches for handling videos with distribution shifts in source and target domains remain source-dependent as they require access to the source data during the adaptation stage. In this paper, we present a self-training based source-free video domain adaptation approach to address this challenge by bridging the gap between the source and the target domains. We use the source pre-trained model to generate pseudo-labels for the target domain samples, which are inevitably noisy. Thus, we treat the problem of source-free video domain adaptation as learning from noisy labels and argue that the samples with correct pseudo-labels can help us in adaptation. To this end, we leverage the cross-entropy loss as an indicator of the correctness of the pseudo-labels and use the resulting small-loss samples from the target domain for fine-tuning the model. We further enhance the adaptation performance by implementing a teacher–student (TS) framework, in which the teacher, which is updated gradually, produces reliable pseudo-labels. Meanwhile, the student undergoes fine-tuning on the target domain videos using these generated pseudo-labels to improve its performance. Extensive experimental evaluations show that our methods, termed as CleanAdapt, CleanAdapt + TS, achieve state-of-the-art results, outperforming the existing approaches on various open datasets.