1167 research outputs found
Sort by
Attribute or Abstain: Large Language Models as Long Document Assistants
No full textThis folder contains data to run experiments on LAB, the Long document Attribution Benchmark introduced in "Attribute or Abstain: Large Language Models as Long Document Assistants". All datasets were converted to the Intertext Graph format introduced by Kuznetsov et al. (2022)
Dataset of mechanical coupling between the upper body and the lower body during walking
No full textThis is a dataset from human walking experiments including ground reaction forces (GRFs) and inertial forces of different body parts. The detailed description can be found in this paper: https://doi.org/10.3389/fnbot.2022.88364
Supplementary data of Resonant Nanopumps: AC Gate Voltages in Conical Nanopores Induce Directed Electrolyte Flow
No full textThis collection contains the raw data, numerical model, and other supporting material for the following publication:
Aaron D. Ratschow, Doyel Pandey, Benno Liebchen, Somnath Bhattacharyya, and Steffen Hardt; ' Resonant Nanopumps: ac Gate Voltages in Conical Nanopores Induce Directed Electrolyte Flow ', Phys. Rev. Lett. 129(26), 264501 (2022).
DOI: https://doi.org/10.1103/PhysRevLett.129.264501
The raw data for the published results are kept under the ''data published'' folder.
The computational finite-element model (Comsol Multiphysics mph-format and the command history as Matlab code) is stored inside the ''numerical modelling'' folder.
The handwritten notes/derivations are saved under the ''analytical modelling'' folder
FEDI Dataset
No full textFEDI is the first task-oriented document-grounded dialogue dataset for learning from demographic information, user emotions and implicit user feedback. In its current version, FEDI consists of 8,852 dialogues, divided into 1,988 feedback-free dialogues, including 326 test dialogues, and 6,864 feedback dialogues (1,716 in four versions, each with one feedback scneario less per dialogue).0.
Research Data Management: A Perspective from the CRC 1194
Get PDFResearch Data Management: A Perspective from the CRC 11941.
Heat Pump Vitocal 200-G
No full textData set of the heat pump vitocal 200-G in the ETA Research Factory supply system.
Anwendung standardisierte Laborumgebung - Rohdaten der Produktionswochen
No full textDer Datensatz enthält die Rohdaten zweier Produktionswochen, die im Rahmen des Forschungsprojektes DiNaPro (BMBF, Förderkennzeichen: 02J20E500) in den Jahren 2023 und 2024 aufgenommen worden sind. Der Datensatz besteht aus zwei Dateien, welche die Rohdaten der beiden Produktionswochen separiert enthalten:
- Zur eindeutigen Identifikation dient die NFC Tag Id in Spalte A.
- Die Spalten B-E enthalten weiterführende Informationen bezüglich der produzierten Varianten.
- Spalte G beinhaltet Informationen, wenn an einem Prozessschritt die Bauteile aufgrund eines Ausschusses abgemeldet worden sind.
- Die entsprechende Prozessschritt-ID findet sich in Spalte H.
- Spalte I enthält den Anmeldezeitstempel am Prozessschritt.
- Spalte J enthält den Abmeldezeitstempel am Prozessschritt.
- Die Spalten L-N beinhalten weiterführende Informationen bezüglich dem Prozessschritt und dem Standort in den drei Lernfabriken des Instituts
Chemical analysis and data logging of groundwater quality from three monitoring wells at the SKEWS Site, Campus Lichtwiese of TU Darmstadt
Get PDFFor the construction of the seasonal crystalline borehole thermal energy storage (SKEWS-project) at the Lichtwiese campus of the Technical University (TU) of Darmstadt, three groundwater monitoring wells were installed for monitoring purposes from 15.02. to 18.02.2022. These wells with depths of 24.5 m (GWM 1 and 2N) and 36.5 m (GWM 3) have filter sections of 6 m to 22 m (GWM 1 and 2N) and 8 m to 32 m (GWM 3) in the upper aquifer (Figure 1 and Table 1). The height of the upper edge of the pipe (Pegeloberkante = POK) changed since the final completion under the surface took place on the 20th of July 2022. The wells are positioned at a distance of approx. 20 m to 40 m in a triangular arrangement. The groundwater flow is south to north. The aquifer is likely fracture based in highly weathered basaltic rock with an estimated hydraulic conductivity of about 10-5 m/s to 10-6 m/s.
Manual groundwater samples were taken twice a week. Automatic data loggers were installed for continuous hourly measurements. The measured depth of the logger in the water must be converted into absolute height values by calibrating it to the measurements on site. Missing readings from the data loggers related to low battery levels. Outliers resulting from battery changes and temporary extractions of the loggers from the well were removed.
From 09.06. to 25.06.2022, the approx. 40 m deep pipes of the 4 geothermal borehole heat exchanger (BHE) were installed using the rotary method. The deep drilling of the three geothermal BHEs up to 750 m depth took place from 18.07. to 25.10.22, using both pneumatic and hydraulic down-the-hole hammers, as well as the rotary method with a clay-based drilling fluid. Significant water inflows below 40 m were observed with the pneumatic hammer, as well as water losses from the upper aquifer with the hydraulic hammer. Due to borehole instabilities and fractured zones, intermediate cementations and the use of clay-freshwater drilling fluid were necessary. Consequently, groundwater fluctuations and changes in the chemical composition were to be expected during this period.
Short-term pumping tests were conducted on 07.06. - 08.06.22, 15.06. – 22.06.22, and on 08.06.23. Furthermore, from 05.10. to 17.11.2023, a Geothermal Response Test (GRT) was conducted on BHE 2 with a heating power of approx. 120 kW.
The dataset consists of tables and figures for manual and automatic sampling. Furthermore, the well configuration and location are included in a separate file
Videos of the Numerical Simulation of Drop Array Impingement onto Smooth, Superheated Walls
No full textVideos of the numerical simulation of drop array impingement onto smooth, superheated walls. The videos show the following cases: drop array impingement with a spacing of 1.18, 2.15, and 3.33. Impact parameters such as initial drop diameter, impact velocity, and wall superheat are 0.93 mm, 0.53 m/s, and 7.3 K, respectively. Dimensionless numbers such as Reynolds, Weber, Bond, and Jakob numbers are 1775, 48.8, 0.4, and 0.09, respectively. The working fluid is perfluorohexane (FC-72) in a pure vapor atmosphere of the same fluid at ambient pressure. The simulations are carried out in OpenFOAM
LiOn-XA: Unsupervised Domain Adaptation via LiDAR-Only Cross-Modal Adversarial Training
No full textIn this paper, we propose LiOn-XA, an unsupervised domain adaptation (UDA) approach that combines LiDAR-Only Cross-Modal (X) learning with Adversarial training for 3D LiDAR point cloud semantic segmentation to bridge the domain gap arising from environmental and sensor setup changes. Unlike existing works that exploit multiple data modalities like point clouds and RGB image data, we address UDA in scenarios where RGB images might not be available and show that two distinct LiDAR data representations can learn from each other for UDA. More specifically, we leverage 3D voxelized point clouds to preserve important geometric structure in combination with 2D projection-based range images that provide information such as object orientations or surfaces. To further align the feature space between both domains, we apply adversarial training using both features and predictions of both 2D and 3D neural networks.
Our experiments on 3 real-to-real adaptation scenarios demonstrate the effectiveness of our approach, achieving new state-of-the-art performance when compared to previous uni- and multi-model UDA methods. Our source code is publicly available at https://github.com/JensLe97/lion-xa