arXiv subjects

Monika Arora

Publications and source records attributed to Monika Arora.

4 recordsLinked to original sources

arXiv2026

Selective Neuron Amplification in Transformer Language Models

Large language models often fail on tasks they seem to already understand. In our experiments, this appears to be less about missing knowledge and more about certain internal circuits not being strongly activated during inference. We explore Selective Neuron Amplification, which increases the influence of task relevant neurons without changing the model's parameters. The method works at inference time and does not permanently alter the model. SNA helps mainly when the model is uncertain, while having low effect when the model is already confident. This suggests that some model failures are due to weak activation rather than lack of capability.

cs.LG↗

arXiv2023

Constraints on the parameters of modified Chaplygin-Jacobi and modified Chaplygin-Abel gases in $f(T)$ gravity

In this study, we investigate two dark energy models, MCJG and MCAG, in the context of $f(T)$ gravity within a non-flat FLRW Universe. Our analysis considers radiation, dark matter, and dark energy components. We compare the equation of state for MCJG and MCAG with $f(T)$ gravity. Using recent astronomical data (e.g., $H(z)$, type Ia supernovae, Gamma Ray Bursts, quasars, and BAO), we constrain the models' parameters and explore the Universe's behavior. The reduced Hubble parameter is expressed in terms of observable parameters like $\Omega_{r0}$, $\Omega_{m0}$, $\Omega_{k0}$, $\Omega_{CJ0}$, $\Omega_{CA0}$, and $H_0$. We investigate cosmic evolution using deceleration, $\mathrm{Om}$, and statefinder diagnostics. Information criteria are employed to assess model viability, comparing against the standard $\Lambda$CDM model. Our objective is to deepen our understanding of dark energy, its relation to $f(T)$ gravity, and the mechanisms governing the accelerated expansion of the Universe.

gr-qc↗

arXiv2021

Symmetry-dependent ultrafast manipulation of nanoscale magnetic domains

Symmetry is a powerful concept in physics, but its applicability to far-from-equilibrium states is still being understood. Recent attention has focused on how far-from-equilibrium states lead to spontaneous symmetry breaking. Conversely, ultrafast optical pumping can be used to drastically change the energy landscape and quench the magnetic order parameter in magnetic systems. Here, we find a distinct symmetry-dependent ultrafast behaviour by use of ultrafast x-ray scattering from magnetic patterns with varying degrees of isotropic and anisotropic symmetry. After pumping with an optical laser, the scattered intensity reveals a radial shift exclusive to the isotropic component and exhibits a faster recovery time from quenching for the anisotropic component. These features arise even when both symmetry components are concurrently measured, suggesting a correspondence between the excitation and the magnetic order symmetry. Our results underline the importance of symmetry as a critical variable to manipulate the magnetic order in the ultrafast regime.

cond-mat.mes-hall↗

arXiv2019

Near-Unity Spin Hall Ratio in Ni$_x$Cu$_{1-x}$ Alloys

We report a large spin Hall effect in the 3$d$ transition metal alloy Ni$_x$Cu$_{1-x}$ for $x\in\left\{ 0.3,0.75\right\} $, detected via the ferromagnetic resonance of a Permalloy (Py = Ni$_{80}$Fe$_{20}$) film deposited in a bilayer with the alloy. A thickness series at $x$ = 0.6, for which the alloy is paramagnetic at room temperature, allows us to determine the spin Hall ratio $θ_{\rm{SH}}\approx1$, spin diffusion length $λ_{\rm{s}}$, spin mixing conductance $G_{\uparrow\downarrow}$, and damping $α_{\rm{SML}}$ due to spin memory loss . We compare our results with similar experiments on Py/Pt bilayers measured using the same method. Ab initio band structure calculations with disorder and spin-orbit coupling suggest an intrinsic spin Hall effect in Ni$_x$Cu$_{1-x}$ alloys, although the experiments here cannot distinguish between extrinsic and intrinsic mechanisms.

cond-mat.mes-hall↗