1167 research outputs found
Sort by
Phi 1.5 Model checkpoints for Fine-Tuning with Divergent Chains of Thought Boosts Reasoning Through Self-Correction in Language Models
Phi 1.5 Model checkpoints for Fine-Tuning with Divergent Chains of Thought Boosts Reasoning Through Self-Correction in Language Models1.
Condensing Boiler Vitodens 200-W B2HA
Data set of the condensing boiler vitodens 200-W B2HA in the ETA Research Factory supply system.
Gewichtung Softwareanforderungen Anwendung standardisierte Laborumgebung
Die vorliegenden Daten umfassen einen Paarvergleich für Softwareanforderungen. Insgesamt sind sieben Expert:innen bzgl. Ihrer Meinung zu der Wichtigkeit der zwölf definierten Anforderungen befragt worden. Die jeweiligen Ergebnisse sind in den Tabellen Person1 - Person7 dargestellt. Die Zusammenfassung und daraus resultierende abschließende Gewichtung findet sich im Tabellenblatt Zusammenfassung
Few-Shot-150T (FS150T) Corpus
The Few-Shot-150T Corpus includes 21,600 sentences over 150 controversial topics.
Each sentence was annotated via crowdsourcing as either a supporting argument, an attacking argument,
or not an argument with respect to the topic. See the readme.md in the archive for a data format description.
The annotations (not the annotated sentences) are released under the Creative Commons Attribution-NonCommercial licence (CC BY-NC)
Flight Simulation Data from Hybrid UAV Missions Affected by Degradation
A flight simulation environment is used to analyse the influence of degradation-related functional losses of the various UAV actuators on flight stability. The increasing wear of individual UAV components with varying degradation rates is modelled for 38 simulated UAVs. The simulation of the degradation of the UAV subsystems includes the BLDC motors, the servo motors and the battery
Frequency-based hopping data
***Human hopping experiment:
The hopping experiments involved six healthy young male subjects (age: 24.1 ± 3.3 years, mass: 73.5 ± 6.6 kg), as approved by the Technical University of Darmstadt's Ethical Committee in line with the Declaration of Helsinki guidelines. Subjects provided written consent before participating. Initially, subjects' preferred hopping frequency (PHF) was established at 2.5 ± 0.38 Hz through a 20-second self-selected frequency and height hopping task. Subsequently, they executed hopping trials at 75%, 100%, 125%, and 150% of their PHF, with six 20-second trials per frequency, guided by an acoustic metronome and spaced by at least two minutes of rest to minimize potential biases like muscle stiffness changes or fatigue.
Kinematics were tracked using 20 reflective markers placed on specific anatomical locations and captured by a 10-camera infrared system (Qualisys, 500 Hz, Sweden). Ground reaction force (GRF) and center of pressure (CoP) data were collected for each leg using two Kistler force plates (1 kHz, Switzerland). These data were processed using OpenSim's inverse kinematics and dynamics tools, with signals low-pass filtered at 50 Hz.
For analysis, 20 hops from three trials per subject and frequency were selected based on minimal variation in kinematic and kinetic data. The mean of these hops represented the trial's data.
For additional details, please refer to the corresponding paper.
*** EPA-Hopper-II hopping experiment:
The EPA-Hopper-II, a successor to previous hopper robots, is a unilegged robot engineered to mimic human-like hopping using a blend of electric and pneumatic actuation. Its design, featuring a 3-segment leg that parallels human leg structure, aims to explore how mechanical design and control contribute to replicating human hopping dynamics across various frequencies. The robot leverages electric motors for hip movement during flight phase adjustments and pneumatic artificial muscles (PAMs) for knee and ankle actuation, reflecting the shift from passive to active control seen in human motion. Constructed primarily from 3D printed materials and lightweight carbon fiber tubes, the robot weighs 3.75 kg and includes a compliant foot design for efficiency and shock absorption. It utilizes direct-drive mechanisms for precise control without needing force/torque sensors, powered by a high-capacity battery. While focusing on leg mechanics, the design omits hip actuation complexities, relying on fixed PAM pressures for simplicity. Ground reaction forces and joint angles are meticulously measured to study the robot's hopping behavior.
For additional details, please refer to the corresponding paper
Acceptance and Trust: Drivers’ First Contact with Released Automated Vehicles in Naturalistic Traffic - Questionnaire Data
Diese Studie untersucht die Auswirkungen des Erstkontakts von Fahrern mit einem SAE Level 3 Automatisierten Fahrsystem(ADS) unter realen Verkehrsbedingungen und konzentriert sich dabei auf den Mercedes-Benz Drive Pilot im EQS. Untersucht werden Akzeptanz, Vertrauen, Benutzerfreundlichkeit und Benutzererfahrung. Die Untersuchung wurde auf einem deutschen Autobahnabschnitt mit 30 Teilnehmern, die mit Level 3 ADS nicht vertraut waren. Vor Fragebögen vor und nach der Fahrt wurden verwendet, um die Veränderungen der Akzeptanz und des Vertrauens. Ergänzende Messgrößen waren Bewertungen der Benutzerfreundlichkeit und der Benutzererfahrung nach der Fahrt
Gearbox Condition Monitoring based on Angle Measurement using a Gear as Material Measure - Test data
Test data that is used in the PhD by Yanik Koch.
The test data was acquired in a series of three test series. The initial test was part of a collaboration project with Sensitec GmbH in 2020. The second test was part of the Master thesis of Marten Schäffner executed in 2022. The third test was part of the Bachelor thesis of Pablo de Orte.
A short description of the test stand with the naming convention of the data is added as PDF. Additional explanations can be found in the PhD thesis by Yanik Koch
Fluid wetting and penetration characteristics in T-shaped microchannels - Data
Measurement data and data-processing scripts accompanying the "Fluid wetting and penetration characteristics in T-shaped microchannels" publication.1.
Evidence - Computer-assisted Interactive Extraction of Dictionary Examples from Large Corpora
Anonymized models from the expert and lay-user studies conducted in the project Evidence. Each model was train for 50-60 iterations on a specific word class (adjective, verb, noun-monoseme, and noun-polyseme)