Ali Zaidi

Neil Byers, Ali Zaidi, Valerie Skye et al.

Efforts have increased in recent years to identify associations between specific datasets and the scientific literature that incorporates them. Knowing that a given publication cites a given dataset, the next logical step is to explore how or why that data was used. Advances in recent years with pretrained, transformer-based large language models (LLMs) offer potential means for scaling the description of data use cases in the published literature. This avoids expensive manual labeling and the development of training datasets for classical machine-learning (ML) systems. In this work we apply an open-source LLM, Llama 3.1-405B, to generate structured data use case labels for publications known to incorporate specific genomic datasets. We also introduce a novel evaluation framework for determining the efficacy of our methods. Our results demonstrate that the stock model can achieve an F1 score of .674 on a zero-shot data citation classification task with no previously defined categories. While promising, our results are qualified by barriers related to data availability, prompt overfitting, computational infrastructure, and the expense required to conduct responsible performance evaluation.

Aleksei Gonnochenko, Aleksandr Semochkin, Dmitry Egorov et al.

Given the laborious difficulty of moving heavy bags of physical currency in the cash center of the bank, there is a large demand for training and deploying safe autonomous systems capable of conducting such tasks in a collaborative workspace. However, the deformable properties of the bag along with the large quantity of rigid-body coins contained within it, significantly increases the challenges of bag detection, grasping and manipulation by a robotic gripper and arm. In this paper, we apply deep reinforcement learning and machine learning techniques to the task of controlling a collaborative robot to automate the unloading of coin bags from a trolley. To accomplish the task-specific process of gripping flexible materials like coin bags where the center of the mass changes during manipulation, a special gripper was implemented in simulation and designed in physical hardware. Leveraging a depth camera and object detection using deep learning, a bag detection and pose estimation has been done for choosing the optimal point of grasping. An intelligent approach based on deep reinforcement learning has been introduced to propose the best configuration of the robot end-effector to maximize successful grasping. A boosted motion planning is utilized to increase the speed of motion planning during robot operation. Real-world trials with the proposed pipeline have demonstrated success rates over 96\% in a real-world setting.