Search

We present two multilingual LLMs designed to embrace Europe's linguistic diversity by supporting all 24 official languages of the European Union. Trained on a dataset comprising around 60% non-English data and utilizing a custom multilingual tokenizer, our models address the limitations of existing LLMs that predominantly focus on English or a few high-resource languages. We detail the models' development principles, i.e., data composition, tokenizer optimization, and training methodologies. The models demonstrate competitive performance across multilingual benchmarks, as evidenced by their performance on European versions of ARC, HellaSwag, MMLU, and TruthfulQA.

A major challenge of the long measurement times in magnetic resonance imaging (MRI), an important medical imaging technology, is that patients may move during data acquisition. This leads to severe motion artifacts in the reconstructed images and volumes. In this paper, we propose a deep learning-based test-time-training method for accurate motion estimation. The key idea is that a neural network trained for motion-free reconstruction has a small loss if there is no motion, thus optimizing over motion parameters passed through the reconstruction network enables accurate estimation of motion. The estimated motion parameters enable to correct for the motion and to reconstruct accurate motion-corrected images. Our method uses 2D reconstruction networks to estimate rigid motion in 3D, and constitutes the first deep learning based method for 3D rigid motion estimation towards 3D-motion-corrected MRI. We show that our method can provably reconstruct motion parameters for a simple signal and neural network model. We demonstrate the effectiveness of our method for both retrospectively simulated motion and prospectively collected real motion-corrupted data.

Recent cutting-edge experiments have provided {\it in situ} structure characterization and measurements of the pressure ($P$), density ($\bar{\rho}$) and temperature ($T$) of shock compressed silicon in the 100 GPa range of pressures and upto $\sim$10,000K. We present first-principles calculations in this $P,T,\bar{\rho}$ regime to reveal a plethora of novel liquid-liquid phase transitions (LPTs) identifiable via discontinuities in the pressure and the compressibility. Evidence for the presence of a highly-correlated liquid (CL) phase, as well as a normal-liquid (NL) phase at the LPTs is presented by a detailed study of one LPT. The LPTs make the interpretation of these experiments more challenging. The LPTs preserve the short-ranged ionic structure of the fluid by collective adjustments of many distant atoms when subject to compression and heating, with minimal change in the ion-ion pair-distribution functions, and in transport properties such as the electrical and thermal conductivities $\sigma$ and $\kappa$. We match the experimental X-Ray Thomson scattering and X-ray diffraction data theoretically, and provide pressure isotherms, ionization data and compressibilities that support the above picture of liquid silicon as a highly complex LPT-driven ``glassy'' metallic liquid. These novel results are relevant to materials research, studies of planetary interiors, high-energy-density physics, and in laser-fusion studies., Comment: 8 pages, six figures

This work presents a cost-effective method for collecting single photons emitted from single nitrogen-vacancy centers in nanodiamonds. Conventional components of a confocal laser-scanning microscope, such as microscope objectives and the piezo translation stages, are replaced by two affordable Blu-ray optical pickup units. A Hanbury Brown and Twiss setup is used to identify single photon emission. The proposed approach is inexpensive and simple and lowers the entry-level to single photon research for quantum technologies. This enables student lab experiments or demonstration experiments at schools and shows that efficient sources of quantum light can be made from standard components compatible with established industry processes.

Reliable control of quantum information in matter-based qubits requires precisely applied external fields, and unaccounted for spatial cross-talk of these fields between adjacent qubits leads to loss of fidelity. We report a CMOS foundry-produced, micro-fabricated silicon nitride (Si3N4) optical waveguide for addressing a chain of eight, unequally-spaced trapped barium ions with crosstalk compatible with scalable quantum information processing. The crosstalk mitigation techniques incorporated into the chip design result in a reduction of the measured optical field by at least 50.8(1.3) dB between adjacent waveguide outputs near 650 nm and similar behavior for devices designed for 493 nm and 585 nm. The waveguide outputs near 650 nm, along with a global laser near 493 nm were used to laser-cool a chain of eight barium-138 ions, and a camera imaged the resulting fluorescence at 493 nm., Comment: 9 pages, 7 figures

The Multimodal Sentiment Analysis Challenge (MuSe) 2024 addresses two contemporary multimodal affect and sentiment analysis problems: In the Social Perception Sub-Challenge (MuSe-Perception), participants will predict 16 different social attributes of individuals such as assertiveness, dominance, likability, and sincerity based on the provided audio-visual data. The Cross-Cultural Humor Detection Sub-Challenge (MuSe-Humor) dataset expands upon the Passau Spontaneous Football Coach Humor (Passau-SFCH) dataset, focusing on the detection of spontaneous humor in a cross-lingual and cross-cultural setting. The main objective of MuSe 2024 is to unite a broad audience from various research domains, including multimodal sentiment analysis, audio-visual affective computing, continuous signal processing, and natural language processing. By fostering collaboration and exchange among experts in these fields, the MuSe 2024 endeavors to advance the understanding and application of sentiment analysis and affective computing across multiple modalities. This baseline paper provides details on each sub-challenge and its corresponding dataset, extracted features from each data modality, and discusses challenge baselines. For our baseline system, we make use of a range of Transformers and expert-designed features and train Gated Recurrent Unit (GRU)-Recurrent Neural Network (RNN) models on them, resulting in a competitive baseline system. On the unseen test datasets of the respective sub-challenges, it achieves a mean Pearson's Correlation Coefficient ($\rho$) of 0.3573 for MuSe-Perception and an Area Under the Curve (AUC) value of 0.8682 for MuSe-Humor.

A better understanding of interactive pedestrian behavior in critical traffic situations is essential for the development of enhanced pedestrian safety systems. Real-world traffic observations play a decisive role in this, since they represent behavior in an unbiased way. In this work, we present an approach of how a subset of very considerable pedestrian-vehicle interactions can be derived from a camera-based observation system. For this purpose, we have examined road user trajectories automatically for establishing temporal and spatial relationships, using 110h hours of video recordings. In order to identify critical interactions, our approach combines the metric post-encroachment time with a newly introduced motion adaption metric. From more than 11,000 reconstructed pedestrian trajectories, 259 potential scenarios remained, using a post-encroachment time threshold of 2s. However, in 95% of cases, no adaptation of the pedestrian behavior was observed due to avoiding criticality. Applying the proposed motion adaption metric, only 21 critical scenarios remained. Manual investigations revealed that critical pedestrian vehicle interactions were present in 7 of those. They were further analyzed and made publicly available for developing pedestrian behavior models3. The results indicate that critical interactions in which the pedestrian perceives and reacts to the vehicle at a relatively late stage can be extracted using the proposed method., Comment: 7 pages, 8 figures, ITSC 2023 accepted

Quantum optimization algorithms (QOAs) have the potential to fundamentally transform the application of optimization methods in decision making. For certain classes of optimization problems, it is widely believed that QOA enables significant run-time performance benefits over current state-of-the-art solutions. With the latest progress on building quantum computers entering the industrialization stage, quantum-based optimization algorithms have become more relevant. The recent extreme increase in the number of publications in the field of QOA demonstrates the growing importance of the topic in both the academia and the industry. The objectives of this paper are as follows: (1) First, we provide insight into the main techniques of quantum-based optimization algorithms for decision making. (2) We describe and compare the two basic classes of adiabatic and gate-based optimization algorithms and argue their potentials and limitations. (3) Herein, we also investigate the key operations research application areas that are expected to be considerably impacted by the use of QOA in decision making in the future. (4) Finally, current implications arising from the future use of QOA from an operations research perspective are discussed., Comment: 16 pages, 6 figures

The recent success of Large Language Models (LLMs) has been predominantly driven by curating the training dataset composition, scaling of model architectures and dataset sizes and advancements in pretraining objectives, leaving tokenizer influence as a blind spot. Shedding light on this underexplored area, we conduct a comprehensive study on the influence of tokenizer choice on LLM downstream performance by training 24 mono- and multilingual LLMs at a 2.6B parameter scale, ablating different tokenizer algorithms and parameterizations. Our studies highlight that the tokenizer choice can significantly impact the model's downstream performance and training costs. In particular, we find that the common tokenizer evaluation metrics fertility and parity are not always predictive of model downstream performance, rendering these metrics a questionable proxy for the model's downstream performance. Furthermore, we show that multilingual tokenizers trained on the five most frequent European languages require vocabulary size increases of factor three in comparison to English. While English-centric tokenizers have been applied to the training of multi-lingual LLMs in the past, we find that this approach results in a severe downstream performance degradation and additional training costs of up to 68%, due to an inefficient tokenization vocabulary.