Search

Future tracking systems in High Energy Physics experiments will require large instrumented areas with low radiation length. Crystalline silicon sensors have been used in tracking systems for decades, but are difficult to manufacture and costly to produce for large areas. We are exploring alternative sensor materials that are amenable to fast fabrication techniques used for thin film devices. Indium Phosphide pad sensors were fabricated at Argonne National Lab using commercially available InP:Fe 2-inch mono-crystal substrates. Current-voltage and capacitance-voltage characterizations were performed to study the basic operating characteristics of a group of sensors. Micro-focused X-ray beams at Canadian Light Source and Diamond Light Source were used to study the response to ionizing radiation, and characterize the uniformity of the response for several devices. The results show a high degree of performance uniformity in our evaluations, both within a device and between the 48 tested devices. This motivates further studies into thin film devices for future tracking detectors., Comment: 30 pages, 27 figures

We investigate the signal propagation in AC-LGAD (aka RSD), which are LGAD with a common N+ layer and segmented AC-coupled readout contacts, by measuring response to IR laser TCT on a large selection of AC-LGAD with strip readout. The interest for this topic derives from the realization that while large charge sharing between neighboring strips is essential for good position resolution, large sharing beyond the next neighbor generates background signals which in general are detrimental to the sensor goal of low occupancy. Using AC-LGAD with strip readout produced by Hamamatsu Photonics (HPK), we evaluate the effects of a variety of sensor properties, including geometrical parameters (strip length, width), process parameters like the N+ layer resistivity, the coupling capacitance, and the thickness of the bulk on the signal sharing and the position resolution., Comment: 9 pages, 7 figures, 13th Hiroshima Symposium HSTD13

The remarkable success of pretrained language models has motivated the study of what kinds of knowledge these models learn during pretraining. Reformulating tasks as fill-in-the-blanks problems (e.g., cloze tests) is a natural approach for gauging such knowledge, however, its usage is limited by the manual effort and guesswork required to write suitable prompts. To address this, we develop AutoPrompt, an automated method to create prompts for a diverse set of tasks, based on a gradient-guided search. Using AutoPrompt, we show that masked language models (MLMs) have an inherent capability to perform sentiment analysis and natural language inference without additional parameters or finetuning, sometimes achieving performance on par with recent state-of-the-art supervised models. We also show that our prompts elicit more accurate factual knowledge from MLMs than the manually created prompts on the LAMA benchmark, and that MLMs can be used as relation extractors more effectively than supervised relation extraction models. These results demonstrate that automatically generated prompts are a viable parameter-free alternative to existing probing methods, and as pretrained LMs become more sophisticated and capable, potentially a replacement for finetuning., Comment: v2: Fixed error in Figure 2

Recent advances in deep neural networks have enabled algorithms to compose music that is comparable to music composed by humans. However, few algorithms allow the user to generate music with tunable parameters. The ability to tune properties of generated music will yield more practical benefits for aiding artists, filmmakers, and composers in their creative tasks. In this paper, we introduce DeepJ - an end-to-end generative model that is capable of composing music conditioned on a specific mixture of composer styles. Our innovations include methods to learn musical style and music dynamics. We use our model to demonstrate a simple technique for controlling the style of generated music as a proof of concept. Evaluation of our model using human raters shows that we have improved over the Biaxial LSTM approach.