Search for: All records

The imbalance of matter and antimatter in our Universe provides compelling motivation to search for undiscovered particles that violate charge-parity symmetry. Interactions with vacuum fluctuations of the fields associated with these new particles will induce an electric dipole moment of the electron (eEDM). We present the most precise measurement yet of the eEDM using electrons confined inside molecular ions, subjected to a huge intramolecular electric field, and evolving coherently for up to 3 seconds. Our result is consistent with zero and improves on the previous best upper bound by a factor of ~2.4. Our results provide constraints on broad classes of new physics above ${10}^{13}$electron volts, beyond the direct reach of the current particle colliders or those likely to be available in the coming decades. 

Abstract. With the greater application of machine learning models in educational contexts, it is important to understand where such meth- ods perform well as well as how they may be improved. As such, it is important to identify the factors that contribute to prediction error in order to develop targeted methods to enhance model accuracy and mitigate risks of algorithmic bias and unfairness. Prior works have led to the development and application of automated assessment methods that leverage machine learning and natural language processing. The performance of these methods have often been reported as being posi- tive, but other prior works have identified aspects on which they may be improved. Particularly in the context of mathematics, the presence of non-linguistic characters and expressions have been identified to con- tribute to observed model error. In this paper, we build upon this prior work by observing a developed automated assessment model for open- response questions in mathematics. We develop a new approach which we call the “Math Term Frequency” (MTF) model to address this issue caused by the presence of non-linguistic terms and ensemble it with the previously-developed assessment model. We observe that the inclusion of this approach notably improves model performance. Finally, we observe how well this ensembled method extrapolates to student responses in the context of Algorithms, a domain similarly characterized by a large number of non-linguistic terms and expressions. This work represents an example of practice of how error analyses can be leveraged to address model limitations.