SearcharxivSearch

arXiv subjects

Minhwan Kim

Publications and source records attributed to Minhwan Kim.

2 recordsLinked to original sources

Median Mishaps between Chirality and Spin-Orbit Torques via Asymmetric Hysteresis

Averaged observations of physical phenomena have been utilized for comprehending specific occurrences in nature; however, these may overlook the crucial characteristics of individual events, thereby leading to diverse conclusions. For example, individual wave occurrences such as superposition and interference yield markedly divergent outcomes when viewed in detail. Similarly, this study enhances the comprehensive framework of spin-orbit torque (SOT) within the hysteresis loop shift measurement by employing the average of effective magnetic fields arising from two distinct magnetic reversals. This approach facilitates the presentation of a physically descriptive SOT model, previously characterized only by single chirality qualitatively. By integrating this model with established measurement methodologies and theoretical paradigms, we advance a theoretical framework based on the magnetic domain-wall chirality of individual polarizations and aim to elucidate the phenomena of SOT with clarity. The anticipated outcomes include the rectification of inaccuracies in widely employed measurement methodologies and the enhancement of our comprehension of the fundamental physics, which are expected to propel advancements in next-generation spintronics materials and devices.

cond-mat.mtrl-sci

The SaTML '24 CNN Interpretability Competition: New Innovations for Concept-Level Interpretability

Interpretability techniques are valuable for helping humans understand and oversee AI systems. The SaTML 2024 CNN Interpretability Competition solicited novel methods for studying convolutional neural networks (CNNs) at the ImageNet scale. The objective of the competition was to help human crowd-workers identify trojans in CNNs. This report showcases the methods and results of four featured competition entries. It remains challenging to help humans reliably diagnose trojans via interpretability tools. However, the competition's entries have contributed new techniques and set a new record on the benchmark from Casper et al., 2023.

cs.LG