arXiv subjects

Namrata Gurung

Publications and source records attributed to Namrata Gurung.

7 recordsLinked to original sources

arXiv2024

Data Bias According to Bipol: Men are Naturally Right and It is the Role of Women to Follow Their Lead

We introduce new large labeled datasets on bias in 3 languages and show in experiments that bias exists in all 10 datasets of 5 languages evaluated, including benchmark datasets on the English GLUE/SuperGLUE leaderboards. The 3 new languages give a total of almost 6 million labeled samples and we benchmark on these datasets using SotA multilingual pretrained models: mT5 and mBERT. The challenge of social bias, based on prejudice, is ubiquitous, as recent events with AI and large language models (LLMs) have shown. Motivated by this challenge, we set out to estimate bias in multiple datasets. We compare some recent bias metrics and use bipol, which has explainability in the metric. We also confirm the unverified assumption that bias exists in toxic comments by randomly sampling 200 samples from a toxic dataset population using the confidence level of 95% and error margin of 7%. Thirty gold samples were randomly distributed in the 200 samples to secure the quality of the annotation. Our findings confirm that many of the datasets have male bias (prejudice against women), besides other types of bias. We publicly release our new datasets, lexica, models, and codes.

cs.CL↗

arXiv2024

Fairness and Bias in Multimodal AI: A Survey

The importance of addressing fairness and bias in artificial intelligence (AI) systems cannot be over-emphasized. Mainstream media has been awashed with news of incidents around stereotypes and other types of bias in many of these systems in recent years. In this survey, we fill a gap with regards to the relatively minimal study of fairness and bias in Large Multimodal Models (LMMs) compared to Large Language Models (LLMs), providing 50 examples of datasets and models related to both types of AI along with the challenges of bias affecting them. We discuss the less-mentioned category of mitigating bias, preprocessing (with particular attention on the first part of it, which we call preuse). The method is less-mentioned compared to the two well-known ones in the literature: intrinsic and extrinsic mitigation methods. We critically discuss the various ways researchers are addressing these challenges. Our method involved two slightly different search queries on two reputable search engines, Google Scholar and Web of Science (WoS), which revealed that for the queries 'Fairness and bias in Large Multimodal Models' and 'Fairness and bias in Large Language Models', 33,400 and 538,000 links are the initial results, respectively, for Scholar while 4 and 50 links are the initial results, respectively, for WoS. For reproducibility and verification, we provide links to the search results and the citations to all the final reviewed papers. We believe this work contributes to filling this gap and providing insight to researchers and other stakeholders on ways to address the challenges of fairness and bias in multimodal and language AI.

cs.CL↗

arXiv2024

Context-based Interpretable Spatio-Temporal Graph Convolutional Network for Human Motion Forecasting

Human motion prediction is still an open problem extremely important for autonomous driving and safety applications. Due to the complex spatiotemporal relation of motion sequences, this remains a challenging problem not only for movement prediction but also to perform a preliminary interpretation of the joint connections. In this work, we present a Context-based Interpretable Spatio-Temporal Graph Convolutional Network (CIST-GCN), as an efficient 3D human pose forecasting model based on GCNs that encompasses specific layers, aiding model interpretability and providing information that might be useful when analyzing motion distribution and body behavior. Our architecture extracts meaningful information from pose sequences, aggregates displacements and accelerations into the input model, and finally predicts the output displacements. Extensive experiments on Human 3.6M, AMASS, 3DPW, and ExPI datasets demonstrate that CIST-GCN outperforms previous methods in human motion prediction and robustness. Since the idea of enhancing interpretability for motion prediction has its merits, we showcase experiments towards it and provide preliminary evaluations of such insights here. available code: https://github.com/QualityMinds/cistgcn

cs.CV↗

arXiv2022

Melting of magnetic order in ${\mathrm{NaOsO}}_{3}$ by fs laser pulses

NaOsO$_3$ has recently attracted significant attention for the strong coupling between its electronic band structure and magnetic ordering. Here, we used time-resolved magnetic X-ray diffraction to determine the timescale of the photoinduced \afm dynamics in NaOsO$_3$. Our measurements are consistent with a sub-100~fs melting of the \afm long-range order, that occurs significantly faster than the lattice dynamics as monitored by the transient change in intensity of selected Bragg structural reflections, which instead show a decrease of intensity on a timescale of several ps.

cond-mat.str-el↗

arXiv2020

Structurally assisted melting of excitonic correlations in 1T-TiSe2

The simultaneous condensation of electronic and structural degrees of freedom gives rise to new states of matter, including superconductivity and charge-density-wave formation. When exciting such a condensed system, it is commonly assumed that the ultrafast laser pulse disturbs primarily the electronic order, which in turn destabilizes the atomic structure. Contrary to this conception, we show here that structural destabilization of few atoms causes melting of the macroscopic ordered charge-density wave in 1T-TiSe2. Using ultrafast pump-probe non-resonant and resonant X-ray diffraction, we observe full suppression of the Se 4p orbital order and the atomic structure at excitation energies more than one order of magnitude below the suggested excitonic binding energy. Complete melting of the charge-density wave occurs 4-5 times faster than expected from a purely electronic charge-screening process, strongly suggesting a structurally assisted breakup of excitonic correlations. Our experimental data clarifies several questions on the intricate coupling between structural and electronic order in stabilizing the charge-density-wave in 1T-TiSe2. The results further show that electron-phonon-coupling can lead to different, energy dependent phase-transition pathways in condensed matter systems, opening new possibilities in the conception of non-equilibrium phenomena at the ultrafast scale.

cond-mat.str-el↗

arXiv2020

Correlation between electronic and structural orders in 1T-TiSe2

The correlation between electronic and crystal structures of 1T-TiSe2 in the charge density wave (CDW) state is studied by x-ray diffraction. Three families of reflections are used to probe atomic displacements and the orbital asymmetry in Se. Two distinct onset temperatures are found, TCDW and a lower T* indicative for an onset of Se out-of-plane atomic displacements. T* coincides with a DC resistivity maximum and the onset of the proposed gyrotropic electronic structure. However, no indication for chirality is found. The relation between the atomic displacements and the transport properties is discussed in terms of Ti 3d and Se 4p states that only weakly couple to the CDW order.

cond-mat.str-el↗

arXiv2020

Ultrafast electron localization in a correlated metal

Ultrafast electron delocalization induced by a fs laser pulse is a well-known process and is the initial step for important applications such as fragmentation of molecules or laser ablation in solids. It is well understood that an intense fs laser pulse can remove several electrons from an atom within its pulse duration. [1] However, the speed of electron localization out of an electron gas, the capture of an electron by ion, is unknown. Here, we demonstrate that electronic localization out of the conduction band can occur within only a few hundred femtoseconds. This ultrafast electron localization into 4f states has been directly quantified by transient x-ray absorption spectroscopy following photo-excitation of a Eu based correlated metal with a fs laser pulse. Our x-ray experiments show that the driving force for this process is either an ultrafast reduction of the energy of the 4f states, a change of their bandwidth or an increase of the hybridization between the 4f and the 3d states. The observed ultrafast electron localization process raises further basic questions for our understanding of electron correlations and their coupling to the lattice.

cond-mat.str-el↗