arXiv subjects

Dhruv Sood

Publications and source records attributed to Dhruv Sood.

3 recordsLinked to original sources

arXiv2026

Optimization of the HHL Algorithm

The Harrow-Hassidim-Lloyd (HHL) algorithm is a quantum algorithm for solving systems of linear equations that, in principle, offers an exponential improvement in scaling with the system size compared to classical approaches. In this work, we investigate the practical implementation and optimisation of the HHL algorithm with a focus on improving its performance on near-term quantum simulators. After outlining the algorithm, we examine two optimisation strategies aimed at improving fidelity and scalability: Suzuki-Trotter decomposition of the Hamiltonian evolution operator and a block-encoding approach that embeds the problem matrix into a larger unitary operator. The performance of these methods is evaluated through simulations on matrices with varying sparsity, including diagonal, tridiagonal, moderately dense, and fully dense cases. Our results show that while HHL achieves near-ideal fidelity for highly structured matrices, performance degrades as sparsity decreases due to the increasing cost of Hamiltonian simulation and reduced post-selection probability due to higher condition number. Block encoding is found to provide improved fidelity for moderately dense matrices, whereas Trotterisation offers a qubit-efficient approach for sparse systems. These results highlight the importance of matrix structure in determining the practical efficiency of HHL and inform future implementations that combine algorithmic optimisation with hardware-aware design.

quant-ph↗

arXiv2025

AMBEDKAR-A Multi-level Bias Elimination through a Decoding Approach with Knowledge Augmentation for Robust Constitutional Alignment of Language Models

Large Language Models (LLMs) can inadvertently reflect societal biases present in their training data, leading to harmful or prejudiced outputs. In the Indian context, our empirical evaluations across a suite of models reveal that biases around caste and religion are particularly salient. Yet, most existing mitigation strategies are Western-centric and fail to address these local nuances. We propose AMBEDKAR, a framework inspired by the egalitarian vision of Dr B. R. Ambedkar, architect of the Indian Constitution, to guide LLM outputs toward fairness, neutrality, and inclusion in line with Articles 14 to 17. Our approach introduces a Constitution-Aware Decoding Layer, guided by the AI Constitution of India and applied only at inference time, without any parameter updates to the base model. We incorporate a speculative decoding algorithm that proactively reduces casteist and communal bias during generation. This mitigation layer operates directly within the decoding process, avoiding changes to model internals and lowering the computational and infrastructural costs associated with retraining. We reinterpret speculative decoding not merely as an efficiency tool but as a mechanism for fairness. In this framework, a Small Language Model (SLM) acts as a potentially biased generator, while a constitutionally guided Large Language Model (LLM) serves as the verifier. Rather than accelerating generation, the LLM enforces bias-robust trajectories in the SLM outputs. This inversion of roles gives rise to a fairness-by-speculation paradigm. Our approach yields an absolute reduction of bias up to 26.41 percent compared to baseline. Our source code, datasets, and results are available at https://anonymous.4open.science/r/AMBEDKAR-983B/

cs.CL↗

arXiv2025

Estimating energy levels from lattice QCD correlation functions using a transfer matrix formalism

We present an efficient method for extracting energy levels from lattice QCD correlation functions by computing the eigenvalues of the transfer matrix associated with the lattice QCD Hamiltonian. While mathematically and numerically equivalent to the recently introduced Lanczos procedure, our approach introduces a novel prescription for removing spurious eigenvalues using a kernel density estimator (KDE) and Gaussian-convoluted histogram method. This strategy yields a robust and stable estimate of the energy spectrum, outperforming the Cullum-Willoughby filtering technique in efficiency. In addition, we detail how this method can be applied to extract overlap factors from two-point correlation functions, as well as matrix elements from three-point functions with a current insertion. Furthermore, we extend the methodology to accommodate correlation matrices constructed from a variational basis of operators, with its Block formulation. We demonstrate the efficacy of this framework by computing the two lowest energy levels for a broad range of hadrons, including several nuclei. Although the signal-to-noise ratio is not significantly improved, the extracted energy levels are found to be more reliable than those obtained with conventional techniques. Within a given statistical ensemble, the proposed method effectively captures both statistical uncertainties and systematic errors, including those arising from the choice of fitting window, making it a robust and practical tool for lattice QCD analysis.

hep-lat↗