arXiv subjects

Minhajul Islam

Publications and source records attributed to Minhajul Islam.

7 recordsLinked to original sources

arXiv2025

Structured Reasoning with Tree-of-Thoughts for Bengali Math Word Problems

Mathematical Word Problems (MWPs) are among the most challenging tasks in natural language processing because they require both linguistic understanding and multi-step numerical reasoning. While Chain-of-Thought (CoT) prompting has shown promise, its linear structure often propagates errors, limiting overall effectiveness. To address this limitation, we present the a systematic study of Tree-of-Thought (ToT) reasoning for Bengali MWPs using the SOMADHAN dataset. Owing to computational and token-cost constraints, we evaluate a curated set of 100 representative problems across multiple large language models (LLMs), including GPT-OSS and LLaMA variants, under standard prompting, CoT, and ToT strategies. Our results show that CoT improves baseline accuracy from 78% (standard prompting) to 83% on average, while ToT further increases performance by up to 5 percentage points, achieving 88% accuracy with GPT-OSS-120B. These improvements highlight that ToT is particularly effective in medium-to-large-scale models but may offer less advantage for smaller ones. Overall, our findings establish ToT as a robust framework for solving mathematical problems in low-resource languages such as Bengali. More broadly, this study shows that structured reasoning methods like ToT can provide more reliable and globally consistent outcomes than CoT, paving the way for better reasoning strategies in multilingual NLP.

cs.CL↗

arXiv2024

Endogenous Crashes as Phase Transitions

This paper explores the mechanisms behind extreme financial events, specifically market crashes, by employing the theoretical framework of phase transitions. We focus on endogenous crashes, driven by internal market dynamics, and model these events as first-order phase transitions critical, stochastic, and dynamic. Through a comparative analysis of early warning signals associated with each type of transition, we demonstrate that dynamic phase transitions (DPT) offer a more accurate representation of market crashes than critical (CPT) or stochastic phase transitions (SPT). Unlike existing models, such as the Log-Periodic Power Law (LPPL) model, which often suffers from overfitting and false positives, our approach grounded in DPT provides a more robust prediction framework. Empirical findings, based on an analysis of S&P 500 stocks from 2019 to 2024, reveal significant trends in volatility and anomalous dimensions before crashes, supporting the superiority of the DPT model. This work contributes to a deeper understanding of the predictive signals preceding market crashes and offers a novel perspective on their underlying dynamics.

q-fin.MF↗

arXiv2023

Unmasking Deepfake Faces from Videos Using An Explainable Cost-Sensitive Deep Learning Approach

Deepfake technology is widely used, which has led to serious worries about the authenticity of digital media, making the need for trustworthy deepfake face recognition techniques more urgent than ever. This study employs a resource-effective and transparent cost-sensitive deep learning method to effectively detect deepfake faces in videos. To create a reliable deepfake detection system, four pre-trained Convolutional Neural Network (CNN) models: XceptionNet, InceptionResNetV2, EfficientNetV2S, and EfficientNetV2M were used. FaceForensics++ and CelebDf-V2 as benchmark datasets were used to assess the performance of our method. To efficiently process video data, key frame extraction was used as a feature extraction technique. Our main contribution is to show the models adaptability and effectiveness in correctly identifying deepfake faces in videos. Furthermore, a cost-sensitive neural network method was applied to solve the dataset imbalance issue that arises frequently in deepfake detection. The XceptionNet model on the CelebDf-V2 dataset gave the proposed methodology a 98% accuracy, which was the highest possible whereas, the InceptionResNetV2 model, achieves an accuracy of 94% on the FaceForensics++ dataset. Source Code: https://github.com/Faysal-MD/Unmasking-Deepfake-Faces-from-Videos-An-Explainable-Cost-Sensitive-Deep-Learning-Approach-IEEE2023

cs.CV↗

arXiv2023

BRST Symmetry of Non-Lorentzian Yang-Mills Theory

We explore the realization of BRST symmetry in the non-Lorentzian Yang-Mills Lagrangian within the context of Galilean and Carrollian Yang-Mills theory. Firstly we demonstrate the nilpotent property of classical BRST transformations and construct corresponding conserve charges for both cases. Then we analyze the algebra of these charges and observe the nilpotent properties at the algebraic level. The findings of this study contribute to a deeper understanding of BRST symmetry in non-Lorentzian Yang-Mills Lagrangians and provide insights into the algebraic properties of related conserve charges.

hep-th↗

arXiv2023

Carrollian Yang-Mills Theory

By doing a small $c$ (speed of light) expansion of $SU(N)$ Yang-Mills fields, we construct two different electric and two different magnetic sectors actions of Carrollian Yang-Mills theory. For both electric and magnetic cases, one sector contains non-trivial self-interaction, and another is $N^2-1$ copies of respective sector Carrollian abelian theory. In $d=4$ , all the four sectors are invariant under infinite Carrollian Conformal symmetry. There are no central extensions when analyzing charge algebra at the phase space level. Lastly, we compute propagators for all four sectors and vertices for two non-trivial sectors. Propagators in position space show ultra-local behavior.

hep-th↗

arXiv2022

Magic Fermions: Carroll and Flat Bands

The Carroll algebra is constructed as the $c\to0$ limit of the Poincare algebra and is associated to symmetries on generic null surfaces. In this paper, we begin investigations of Carrollian fermions or fermions defined on generic null surfaces. Due to the availability of two different (degenerate) metrics on Carroll spacetimes, there is the possibility of two different versions of Carroll Clifford algebras. We consider both possibilities and construct explicit representations of Carrollian gamma matrices and show how to build higher spacetime dimensional representations out of lower ones. Actions for Carroll fermions are constructed with these gamma matrices and the properties of these actions are investigated. We show that in condensed matter systems where the dispersion relation becomes trivial i.e. the energy is not dependent on momentum and bands flatten out, Carroll symmetry generically appears. We give explicit examples of this including that of twisted bi-layer graphene, where superconductivity appears at so called magic angles and connect this to Carroll fermions.

hep-th↗

arXiv2022

Galilean Gauge Theories from Null Reductions

The procedure of null reduction provides a concrete way of constructing field theories with Galilean invariance. We use this to examine Galilean gauge theories, viz. Galilean electrodynamics and Yang-Mills theories in spacetime dimensions 3 and 4. Different non-relativistic conformal symmetries arise in these contexts: Schr{ö}dinger symmetry in $d=3$ and Galilean conformal symmetry in $d=4$. A canonical analysis further reveals that the symmetries enhance to their infinite dimensional versions in phase space and pick up central extensions. In addition, for the Abelian theory, we discuss non-relativistic electro-magnetic duality in $d=3$ and its difference with the $d=4$ version. We also mention some quantum aspects for both Abelian and non-Abelian theories.

hep-th↗