Search

The problem of avoiding saturation of the coil currents is critical in large tokamaks with superconducting coils like ITER. Indeed, if the current limits are reached, a loss of control of the plasma may lead to a major disruption. Therefore, a Current Limit Avoidance (CLA) system is essential to operate safely. This paper provides the first experimental evidence that the online solution of a constrained quadratic optimization problem can offer a valid methodology to implement a CLA. Experiments are carried out on the Tokamak \`a Configuration Variable (TCV) at the Swiss Plasma Center, showing the effectiveness of the proposed approach and its suitability for real-time application in view of future reactors such as ITER., Comment: 14 pages, currently under review

The OpenUniverse2024 simulation suite is a cross-collaboration effort to produce matched simulated imaging for multiple surveys as they would observe a common simulated sky. Both the simulated data and associated tools used to produce it are intended to uniquely enable a wide range of studies to maximize the science potential of the next generation of cosmological surveys. We have produced simulated imaging for approximately 70 deg$^2$ of the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST) Wide-Fast-Deep survey and the Nancy Grace Roman Space Telescope High-Latitude Wide-Area Survey, as well as overlapping versions of the ELAIS-S1 Deep-Drilling Field for LSST and the High-Latitude Time-Domain Survey for Roman. OpenUniverse2024 includes i) an early version of the updated extragalactic model called Diffsky, which substantially improves the realism of optical and infrared photometry of objects, compared to previous versions of these models; ii) updated transient models that extend through the wavelength range probed by Roman and Rubin; and iii) improved survey, telescope, and instrument realism based on up-to-date survey plans and known properties of the instruments. It is built on a new and updated suite of simulation tools that improves the ease of consistently simulating multiple observatories viewing the same sky. The approximately 400 TB of synthetic survey imaging and simulated universe catalogs are publicly available, and we preview some scientific uses of the simulations.

Plasma-terminating disruptions represent a critical outstanding issue for reactor-relevant tokamaks. ITER will use shattered pellet injection (SPI) as its disruption mitigation system to reduce heat loads, vessel forces, and to suppress the formation of runaway electrons. In this paper we demonstrate that reduced kinetic modelling of SPI is capable of capturing the major experimental trends in ASDEX Upgrade SPI experiments, such as dependence of the radiated energy fraction on neon content, or the current quench dynamics. Simulations are also consistent with the experimental observation of no runaway electron generation with neon and mixed deuterium-neon pellet composition. We also show that statistical variations in the fragmentation process only have a notable impact on disruption dynamics at intermediate neon doping, as was observed in experiments.

We present a model for the particle balance in the post-disruption runaway electron plateau phase of a tokamak discharge. The model is constructed with the help of, and applied to, experimental data from TCV discharges investigating the so-called "low-Z benign termination" runaway electron mitigation scheme. In the benign termination scheme, the free electron density is first reduced in order for a subsequently induced MHD instability to grow rapidly and spread the runaway electrons widely across the wall. The model explains why there is an upper limit for the neutral pressure above which the termination is not benign. We are also able to show that the observed non-monotonic dependence of the free electron density with the measured neutral pressure is due to plasma re-ionization induced by runaway electron impact ionization. At higher neutral pressures, more target particles are present in the plasma for runaway electrons to collide with and ionize. Parameter scans are conducted to clarify the role of the runaway electron density and energy on the upper pressure limit, and it is found that only the runaway electron density has a noticeable impact., Comment: 11 pages, 8 figures, special issue paper after the Joint Varenna-Lausanne International Workshop

Shattered pellet injection (SPI) is selected for the disruption mitigation system in ITER, due to deeper penetration, expected assimilation efficiency and prompt material delivery. This article describes non-linear simulations of SPI in the ASDEX Upgrade tokamak to test the mitigation efficiency of different injection parameters for neon-doped deuterium pellets using the JOREK code. The simulations are executed as fluid simulations. Additional marker particles are used to evolve the charge state distribution of impurities based on OpenADAS atomic data, i.e., no coronal equilibrium assumption is made. Regarding the pellet composition, neon fraction scans from 0 - 10% are performed. Numerical results show that the thermal quench (TQ) occurs in two stages. In the first stage, approximately half of the thermal energy is lost abruptly, primarily through convective and conductive transport in the stochastic fields. This stage is relatively independent of the injection parameters. In the second stage, where the majority of the remaining thermal energy is lost, radiation plays a dominant role. In cases of very low neon content, this second stage may not occur at all. A larger fraction ($\sim $20%) of the total material in the pellet is assimilated in the plasma for low neon fraction pellets (0.12%) since the full thermal collapse of the plasma occurs later than in high neon fraction scenarios. Nevertheless, the total number of assimilated neon atoms increases with increasing neon fraction. The effects of fragment size and penetration speed are further studied. Slower and smaller fragments promote edge cooling and the formation of a cold front. Faster fragments result in shorter TQ duration and higher assimilation as they reach the hotter plasma regions quicker. Using synthetic diagnostics, comparisons of general trend between simulations and experiments are conducted.

In this paper, the turbulent transport properties of ST40 hot ion plasmas are examined and fully predictive time evolving modelling of a hot ion plasma pulse was performed. Understanding turbulent transport on spherical tokamaks (STs) is challenging due to their unique geometry characteristics. ST40 hot ion plasmas are typically unstable to ion scale Trapped Electron Modes (TEMs) and Ubiquitous Modes (UMs), driven from the kinetic response of trapped particles and passing ions, and electron scale Electron Temperature Gradient Modes (ETGs) at the edge of the plasma. A comparison between the linear unstable modes of the gyro-kinetic code GS2 and the gyro-fluid code TGLF showed that both models agree to a satisfactory level. However, some discrepancy was observed at the core of the plasma where a large fraction of beams ions exists, and electromagnetic effects are potentially important. Turbulent fluxes were also observed to be somewhat overpredicted with TGLF. The core heat ion transport is observed to be close to neoclassical levels due to turbulence suppression from high rotation and fast ion stabilisation, while the edge region is dominated by anomalous transport in both ions and electrons. As a result, enhanced energy confinement is observed in those plasmas driven by the reduced turbulent core region and the confined beam ions. Fully predictive simulations using the ASTRA transport solver coupled with SPIDER, NUBEAM, NCLASS and TGLF together with a novel reduced scrape of layer (SOL) model for the simulation of the last closed flux surface (LCFS) boundary conditions was attempted. Agreement in global quantities but also kinetic profiles between the predictive and interpretative modelling as well as experimental measurements was observed.

XuanLong-50 (EXL-50) is the first medium-size spherical torus (ST) in China, with the toroidal field at major radius at 50 cm around 0.5T. CS-free and non-inductive current drive via electron cyclotron resonance heating (ECRH) was the main physics research issue for EXL-50. Discharges with plasma currents of 50 kA - 180 kA were routinely obtained in EXL-50, with the current flattop sustained for up to or beyond 2 s. The current drive effectiveness on EXL-50 was as high as 1 A/W for low-density discharges using 28GHz ECRH alone for heating power less than 200 kW. The plasma current reached Ip>80 kA for high-density (5*10e18m-2) discharges with 150 kW 28GHz ECRH. Higher performance discharge (Ip of about 120 kA and core density of about 1*10e19m-3) was achieved with 150 kW 50GHz ECRH. The plasma current in EXL-50 was mainly carried by the energetic electrons.Multi-fluid equilibrium model has been successfully applied to reconstruct the magnetic flux surface and the measured plasma parameters of the EXL-50 equilibrium. The physics mechanisms for the solenoid-free ECRH current drive and the energetic electrons has also been investigated. Preliminary experimental results show that 100 kW of lower hybrid current drive (LHCD) waves can drive 20 kA of plasma current. Several boron injection systems were installed and tested in EXL-50, including B2H6 gas puffing, boron powder injection, boron pellet injection. The research plan of EXL-50U, which is the upgrade machine of EXL-50, is also presented.

This paper presents BOUQuET, a multicentric and multi-register/domain dataset and benchmark, and its broader collaborative extension initiative. This dataset is handcrafted in non-English languages first, each of these source languages being represented among the 23 languages commonly used by half of the world's population and therefore having the potential to serve as pivot languages that will enable more accurate translations. The dataset is specially designed to avoid contamination and be multicentric, so as to enforce representation of multilingual language features. In addition, the dataset goes beyond the sentence level, as it is organized in paragraphs of various lengths. Compared with related machine translation (MT) datasets, we show that BOUQuET has a broader representation of domains while simplifying the translation task for non-experts. Therefore, BOUQuET is specially suitable for the open initiative and call for translation participation that we are launching to extend it to a multi-way parallel corpus to any written language.

The COVID-19 pandemic highlighted the urgent need for robust systems to enable rapid data collection, integration, and analysis for public health responses. Existing approaches often relied on disparate, non-interoperable systems, creating bottlenecks in comprehensive analyses and timely decision-making. To address these challenges, the U.S. National Institutes of Health (NIH) launched the Rapid Acceleration of Diagnostics (RADx) initiative in 2020, with the RADx Data Hub, a centralized repository for de-identified and curated COVID-19 data, as its cornerstone. The RADx Data Hub hosts diverse study data, including clinical data, testing results, smart sensor outputs, self-reported symptoms, and information on social determinants of health. Built on cloud infrastructure, the RADx Data Hub integrates metadata standards, interoperable formats, and ontology-based tools to adhere to the FAIR (Findable, Accessible, Interoperable, Reusable) principles for data sharing. Initially developed for COVID-19 research, its architecture and processes are adaptable to other scientific disciplines. This paper provides an overview of the data hosted by the RADx Data Hub and describes the platform's capabilities and architecture.

We present the first experimental measurements of the toroidal mode number of shear Alfv\'en waves in the TJ-II stellarator. A series of experiments were carried out in three different magnetic configurations to investigate counter-NBI driven modes. Co- and counter- electron-cyclotron current drive was used to modify the rotational transform ($\iota/2\pi$) profile leading to the destabilization of a varied set of Alfv\'en eigenmodes with different frequencies and mode numbers. To characterize the spatial structure of the modes we have used two Mirnov probe arrays, one dedicated to the measurement of the poloidal mode number and the other, a recently commissioned helical tri-axial array, dedicated to the measurement of the toroidal mode number. A heavy ion beam probe, operated in radial sweep mode, was employed to characterize the radial location of the modes. We show that the induced changes in $\iota/2\pi$, that are fundamental when it comes to validation studies, cannot be measured experimentally with motional Stark effect so, instead, the shielding current diffusion equation is solved in cylindrical geometry to estimate these changes. We calculate the incompressible shear Alfv\'en continuum for selected cases using \texttt{STELLGAP} and find reasonable consistency with observations. A database with the observed modes has been created, so that it can be used in future work for theory validation purposes.

Language model pretraining with next token prediction has proved effective for scaling compute but is limited to the amount of available training data. Scaling reinforcement learning (RL) unlocks a new axis for the continued improvement of artificial intelligence, with the promise that large language models (LLMs) can scale their training data by learning to explore with rewards. However, prior published work has not produced competitive results. In light of this, we report on the training practice of Kimi k1.5, our latest multi-modal LLM trained with RL, including its RL training techniques, multi-modal data recipes, and infrastructure optimization. Long context scaling and improved policy optimization methods are key ingredients of our approach, which establishes a simplistic, effective RL framework without relying on more complex techniques such as Monte Carlo tree search, value functions, and process reward models. Notably, our system achieves state-of-the-art reasoning performance across multiple benchmarks and modalities -- e.g., 77.5 on AIME, 96.2 on MATH 500, 94-th percentile on Codeforces, 74.9 on MathVista -- matching OpenAI's o1. Moreover, we present effective long2short methods that use long-CoT techniques to improve short-CoT models, yielding state-of-the-art short-CoT reasoning results -- e.g., 60.8 on AIME, 94.6 on MATH500, 47.3 on LiveCodeBench -- outperforming existing short-CoT models such as GPT-4o and Claude Sonnet 3.5 by a large margin (up to +550%)., Comment: 25 pages

We present Fanar, a platform for Arabic-centric multimodal generative AI systems, that supports language, speech and image generation tasks. At the heart of Fanar are Fanar Star and Fanar Prime, two highly capable Arabic Large Language Models (LLMs) that are best in the class on well established benchmarks for similar sized models. Fanar Star is a 7B (billion) parameter model that was trained from scratch on nearly 1 trillion clean and deduplicated Arabic, English and Code tokens. Fanar Prime is a 9B parameter model continually trained on the Gemma-2 9B base model on the same 1 trillion token set. Both models are concurrently deployed and designed to address different types of prompts transparently routed through a custom-built orchestrator. The Fanar platform provides many other capabilities including a customized Islamic Retrieval Augmented Generation (RAG) system for handling religious prompts, a Recency RAG for summarizing information about current or recent events that have occurred after the pre-training data cut-off date. The platform provides additional cognitive capabilities including in-house bilingual speech recognition that supports multiple Arabic dialects, voice and image generation that is fine-tuned to better reflect regional characteristics. Finally, Fanar provides an attribution service that can be used to verify the authenticity of fact based generated content. The design, development, and implementation of Fanar was entirely undertaken at Hamad Bin Khalifa University's Qatar Computing Research Institute (QCRI) and was sponsored by Qatar's Ministry of Communications and Information Technology to enable sovereign AI technology development.

The phenomena of island healing and configuration transition induced by high-power electron cyclotron resonance heating (ECRH) have been investigated in the island divertor configuration on the J-TEXT tokamak. Experimental results reveal that the size of the edge open magnetic island with mode number m/n = 3/1 decreases substantially under specific ECRH conditions. This process, referred to as island healing, occurs when ECRH with a power of 500~600 kW is deposited in the plasma core or when 250 kW of ECRH is deposited at r = 0.5 a, where a is the minor radius. The reduction of the island width makes the island divertor ineffective and transition into the limiter configuration. A model incorporating the influence of ECRH on the scrape-off layer (SOL) thermoelectric current is proposed to explain the observed changes in the edge magnetic topology of the island divertor configuration. These findings suggest that ECRH should be deposited at the plasma core with carefully controlled power to ensure the stable and compatible operation of ECRH and the island divertor configuration in tokamaks. The results can provide insights into achieving robust operation of an island divertor in tokamaks.

Neutrino telescopes, an extension of traditional multiwavelength astronomy, provide a complementary view of the universe using neutrinos. Differences in detector geometry and detection medium mean that improvements to reconstruction techniques made at one experiment are not readily applicable to another. Recently, deep learning has been shown to improve prediction speed and accuracy and offer indifference to detector geometry and detection medium, providing a unique opportunity for collaboration. This work introduces GraphNeT 2.0, an open-source, detector-agnostic deep learning library for neutrino telescopes and related experiments. GraphNeT enables inter-experimental collaboration on the use and development of advanced methods based on major deep learning paradigms like transformers, normalizing flows, graph neural networks, and more., Comment: Submitted and accepted to ML4PS @ NeurIPS

We present OLMo 2, the next generation of our fully open language models. OLMo 2 includes dense autoregressive models with improved architecture and training recipe, pretraining data mixtures, and instruction tuning recipes. Our modified model architecture and training recipe achieve both better training stability and improved per-token efficiency. Our updated pretraining data mixture introduces a new, specialized data mix called Dolmino Mix 1124, which significantly improves model capabilities across many downstream task benchmarks when introduced via late-stage curriculum training (i.e. specialized data during the annealing phase of pretraining). Finally, we incorporate best practices from T\"ulu 3 to develop OLMo 2-Instruct, focusing on permissive data and extending our final-stage reinforcement learning with verifiable rewards (RLVR). Our OLMo 2 base models sit at the Pareto frontier of performance to compute, often matching or outperforming open-weight only models like Llama 3.1 and Qwen 2.5 while using fewer FLOPs and with fully transparent training data, code, and recipe. Our fully open OLMo 2-Instruct models are competitive with or surpassing open-weight only models of comparable size, including Qwen 2.5, Llama 3.1 and Gemma 2. We release all OLMo 2 artifacts openly -- models at 7B and 13B scales, both pretrained and post-trained, including their full training data, training code and recipes, training logs and thousands of intermediate checkpoints. The final instruction model is available on the Ai2 Playground as a free research demo., Comment: Model demo available at playground.allenai.org

Today's generative AI systems are tuned to present information by default rather than engage users in service of learning as a human tutor would. To address the wide range of potential education use cases for these systems, we reframe the challenge of injecting pedagogical behavior as one of \textit{pedagogical instruction following}, where training and evaluation examples include system-level instructions describing the specific pedagogy attributes present or desired in subsequent model turns. This framing avoids committing our models to any particular definition of pedagogy, and instead allows teachers or developers to specify desired model behavior. It also clears a path to improving Gemini models for learning -- by enabling the addition of our pedagogical data to post-training mixtures -- alongside their rapidly expanding set of capabilities. Both represent important changes from our initial tech report. We show how training with pedagogical instruction following produces a LearnLM model (available on Google AI Studio) that is preferred substantially by expert raters across a diverse set of learning scenarios, with average preference strengths of 31\% over GPT-4o, 11\% over Claude 3.5, and 13\% over the Gemini 1.5 Pro model LearnLM was based on.

The tearing mode, a large-scale MHD instability in tokamak, typically disrupts the equilibrium magnetic surfaces, leads to the formation of magnetic islands, and reduces core electron temperature and density, thus resulting in significant energy losses and may even cause discharge termination. This process is unacceptable for ITER. Therefore, the accurate identification of a magnetic island in real time is crucial for the effective control of the tearing mode in ITER in the future. In this study, based on the characteristics induced by tearing modes, an attention-aware convolutional neural network (AM-CNN) is proposed to identify the presence of magnetic islands in tearing mode discharge utilizing the data from ECE diagnostics in the EAST tokamak. A total of 11 ECE channels covering the range of core is used in the tearing mode dataset, which includes 2.5*10^9 data collected from 68 shots from 2016 to 2021 years. We split the dataset into training, validation, and test sets (66.5%, 5.7%, and 27.8%), respectively. An attention mechanism is designed to couple with the convolutional neural networks to improve the capability of feature extraction of signals. During the model training process, we utilized adaptive learning rate adjustment and early stopping mechanisms to optimize performance of AM-CNN. The model results show that a classification accuracy of 91.96% is achieved in tearing mode identification. Compared to CNN without AM, the attention-aware convolutional neural networks demonstrate great performance across accuracy, recall metrics, and F1 score. By leveraging the deep learning model, which incorporates a physical understanding of the tearing process to identify tearing mode behaviors, the combination of physical mechanisms and deep learning is emphasized, significantly laying an important foundation for the future intelligent control of tearing mode dynamics.

The X-Ray Imaging and Spectroscopy Mission (XRISM) is the 7th Japanese X-ray observatory, whose development and operation are in collaboration with universities and research institutes in Japan, U.S., and Europe, including JAXA, NASA, and ESA. The telemetry data downlinked from the satellite are reduced to scientific products by the pre-pipeline (PPL) and pipeline (PL) software running on standard Linux virtual machines on the JAXA and NASA sides, respectively. We ported the PPL to the JAXA "TOKI-RURI" high-performance computing (HPC) system capable of completing $\simeq 160$ PPL processes within 24 hours by utilizing the container platform of Singularity and its "--bind" option. In this paper, we briefly show the data processing in XRISM and present our porting strategy of PPL to the HPC environment in detail., Comment: 4 pages, 2 figures, proceedings of the Astronomical Data Analysis Software and Systems XXXIV

LLMs have revolutionized the field of artificial intelligence and have emerged as the de-facto tool for many tasks. The current established technology of LLMs is to process input and generate output at the token level. This is in sharp contrast to humans who operate at multiple levels of abstraction, well beyond single words, to analyze information and to generate creative content. In this paper, we present an attempt at an architecture which operates on an explicit higher-level semantic representation, which we name a concept. Concepts are language- and modality-agnostic and represent a higher level idea or action in a flow. Hence, we build a "Large Concept Model". In this study, as proof of feasibility, we assume that a concept corresponds to a sentence, and use an existing sentence embedding space, SONAR, which supports up to 200 languages in both text and speech modalities. The Large Concept Model is trained to perform autoregressive sentence prediction in an embedding space. We explore multiple approaches, namely MSE regression, variants of diffusion-based generation, and models operating in a quantized SONAR space. These explorations are performed using 1.6B parameter models and training data in the order of 1.3T tokens. We then scale one architecture to a model size of 7B parameters and training data of about 2.7T tokens. We perform an experimental evaluation on several generative tasks, namely summarization and a new task of summary expansion. Finally, we show that our model exhibits impressive zero-shot generalization performance to many languages, outperforming existing LLMs of the same size. The training code of our models is freely available., Comment: 49 pages

This paper introduces STRIPE (Simulated Transport of RF Impurity Production and Emission), an advanced modeling framework designed to analyze material erosion and the global transport of eroded impurities originating from radio-frequency (RF) antenna structures in magnetic confinement fusion devices. STRIPE integrates multiple computational tools, each addressing different levels of physics fidelity: SolEdge3x for scrape-off-layer plasma profiles, COMSOL for 3D RF rectified voltage fields, RustBCA code for erosion yields and surface interactions, and GITR for 3D ion energy-angle distributions and global impurity transport. The framework is applied to an ion cyclotron RF heated, L-mode discharge #57877 in the WEST Tokamak, where it predicts a tenfold increase in tungsten erosion at RF antenna limiters under RF-sheath rectification conditions, compared to cases with only a thermal sheath. Highly charged oxygen ions (O6+ and higher) emerge as dominant contributors to tungsten sputtering at the antenna limiters. To verify model accuracy, a synthetic diagnostic tool based on inverse photon efficiency or S/XB coefficients from the ColRadPy-collisional radiative model enables direct comparisons between simulation results and experimental spectroscopic data. Model predictions, assuming plasma composition of 1% oxygen and 99% deuterium, align closely with measured neutral tungsten (W-I) spectroscopic data for the discharge #57877, validating the framework's accuracy. Currently, the STRIPE framework is being extended to investigate plasma-material interactions in other RF-heated linear and toroidal devices, offering valuable insights for RF antenna design, impurity control, and performance optimization in future fusion reactors.

Infrared observations with JWST open up a new window into the chemical composition of the gas in the inner disk (1. Together these data point to a rich chemistry in the inner regions that is linked to the physical structure of these disks (e.g., dust traps) and that may be affected by processes such as radial drift of icy pebbles from the outer to the inner disk., Comment: 6 pages, 2 figures. To appear in the proceedings of IAU Symposium 393, Planetary Science and Exoplanets in the era of JWST, Z. Benkhaldoun, A. Szentgyorgyi, Y. Moulane, eds. (Cambridge University Press)

This study presents ion temperature (\(T_i\)) measurements in the MAST-U divertor, using a Retarding Field Energy Analyzer (RFEA). Steady state measurements were made during an L-Mode plasma with the strike point on the RFEA. ELM measurements were made with the strike point swept over the RFEA. The scenarios are characterized by a plasma current (\(I_p\)) of 750 kA, line average electron density (\(n_e\)) between \(1.6 \times 10^{19}\) and \(4.5 \times 10^{19}\,\text{m}^{-3}\), and Neutral Beam Injection (NBI) power ranging from 1.1 MW to 1.6 MW. The ion temperatures, peaking at approximately 10 eV in steady state, were compared with electron temperatures (\(T_e\)) obtained from Langmuir probes (LP) at the same radial positions. Preliminary findings reveal a \(T_i/T_e\) ratio in the divertor region less than 1 for shot 48008. High temporal resolution measurements captured the dynamics of Edge Localized Modes (ELMs) Burn Through, providing \(T_i\) data as a radial distance from the probe peaking around 20 eV.

Future large tokamaks will operate at high plasma currents and high stored plasma energies. To ensure machine protection in case of a sudden loss of plasma confinement (major disruption), a large fraction of the magnetic and thermal energy must be radiated to reduce thermal loads. The disruption mitigation system for ITER is based on massive material injection in the form of shattered pellet injection (SPI). To support ITER, a versatile SPI system was installed at the tokamak ASDEX Upgrade (AUG). The AUG SPI features three independent pellet generation cells and guide tubes, and each was equipped with different shatter heads for the 2022 experimental campaign. We dedicated over 200 plasma discharges to the study of SPI plasma termination, and in this manuscript report on the results of bolometry (total radiation) analysis. The amount of neon inside the pellets is the dominant factor determining the radiated energy fraction (frad). Large and fast fragments, produced by the $12.5^\circ$ rectangular shatter head, lead to somewhat higher values of frad compared to the $25^\circ$ circular or rectangular heads. This effect is strongest for neon content of $< 4\times10^{20}$ neon atoms injected, where a higher normal velocity component (larger fragments) seems slightly beneficial. While full-sized, 8 mm diameter, pure deuterium (D2) pellets lead to a disruption, the 4 mm or shortened 8 mm pellets of pure D2 did not lead to a disruption. The disruption threshold for pure D2 is found to be around $1\times10^{22}$ deuterium molecules inside the pellet. While the radiated energy fraction of non-disruptive SPI is below 20%, this is increased to 40% during the TQ and VDE phase of the disruptive injections. For deuterium-neon-mix pellets, frad values of $< 90$% are observed, and the curve saturates around 80% for 10% neon mixed into the 8 mm pellets ($2\times10^{21}$ neon atoms)., Comment: 15 pages, 9 figures

2D electron density profiles obtained from coherence imaging spectroscopy in different MAST-U divertor conditions are compared. The data includes variations of strike point position, core electron density, and heating power. The improved performance of the long-legged divertors results in a lower electron density and particle flux at the target compared to configurations with smaller strike point major radius, while also being characterized by lower temperatures and deeper detachment. Comparisons against SOLPS simulations generally show good agreement in profile shape along and across the separatrix. The peaking of the electron density downstream of the detachment front is associated with significant neutral drag acting on the plasma flow.

As the capabilities of code large language models (LLMs) continue to expand, their applications across diverse code intelligence domains are rapidly increasing. However, most existing datasets only evaluate limited application domains. To address this gap, we have developed a comprehensive code evaluation dataset FullStack Bench focusing on full-stack programming, which encompasses a wide range of application domains (e.g., basic programming, data analysis, software engineering, mathematics, and machine learning). Besides, to assess multilingual programming capabilities, in FullStack Bench, we design real-world instructions and corresponding unit test cases from 16 widely-used programming languages to reflect real-world usage scenarios rather than simple translations. Moreover, we also release an effective code sandbox execution tool (i.e., SandboxFusion) supporting various programming languages and packages to evaluate the performance of our FullStack Bench efficiently. Comprehensive experimental results on our FullStack Bench demonstrate the necessity and effectiveness of our FullStack Bench and SandboxFusion., Comment: 26 pages

The Surface Photometry and Accurate Rotation Curves (SPARC) database has provided the community with mass models for 175 nearby galaxies, allowing different research teams to test different dark matter models, galaxy evolution models, and modified gravity theories. Extensive tests, however, are hampered by the somewhat heterogeneous nature of the HI rotation curves and the limited sample size of SPARC. To overcome these limitations, we are working on BIG-SPARC, a new database that consists of about 4000 galaxies with HI datacubes from public telescope archives (APERTIF, ASKAP, ATCA, GMRT, MeerKAT, VLA, and WSRT) and near infrared photometry from WISE. For these galaxies, we will provide homogeneously derived HI rotation curves, surface brightness profiles, and mass models. BIG-SPARC is expected to increase the size of its predecessor by a factor of more than 20. This is a necessary step to prepare for the additional order of magnitude increase in sample size expected from ongoing and future HI surveys with the Square Kilometre Array (SKA) and its pathfinders, Comment: To be published in the proceedings of IAU Symposium 392 "Neutral Hydrogen in and around Galaxies in the SKA Era", 6 pages, 3 figures, 1 table

The Space Telescope Imaging Product Simulator (STIPS) is a Python-based package that can be used to simulate scenes from the upcoming \textit{Nancy Grace Roman Space Telescope} (\nancy). STIPS is able to generate post-pipeline astronomical images of any number of sensor chip assembly (SCA) detectors, up to the entire 18-SCA Wide-Field Instrument array on \nancy. STIPS can inject either point spread functions generated with {\tt WebbPSF}, or extended sources in any of the \nancy filters. The output images can include flat field, dark current, and cosmic ray residuals. Additionally, STIPS includes an estimate of Poisson and readout noise, as well as an estimate of the zodiacal background and internal background from the telescope. However, STIPS does not include instrument saturation, non-linearity, or distortion effects. STIPS is provided as an open source repository on GitHub., Comment: Accepted to PASP. 10 pages, 11 figures

A major challenge in nuclear fusion research is the coherent combination of data from heterogeneous diagnostics and modelling codes for machine control and safety as well as physics studies. Measured data from different diagnostics often provide information about the same subset of physical parameters. Additionally, information provided by some diagnostics might be needed for the analysis of other diagnostics. A joint analysis of complementary and redundant data allows, e.g., to improve the reliability of parameter estimation, to increase the spatial and temporal resolution of profiles, to obtain synergistic effects, to consider diagnostics interdependencies and to find and resolve data inconsistencies. Physics-based modelling and parameter relationships provide additional information improving the treatment of ill-posed inversion problems. A coherent combination of all kind of available information within a probabilistic framework allows for improved data analysis results. The concept of Integrated Data Analysis (IDA) in the framework of Bayesian probability theory is outlined and contrasted with conventional data analysis. Components of the probabilistic approach are summarized and specific ingredients beneficial for data analysis at fusion devices are discussed.

We present flux measurements of Uranus observed at phase angles of 43.9{\deg}, 44.0{\deg}, and 52.4{\deg} by the Multispectral Visible Imaging Camera (MVIC) on the New Horizons spacecraft during 2023, 2010, and 2019, respectively. New Horizons imaged Uranus at a distance of about 24-70 AU (2023) in four color filters, with bandpasses of 400-550 nm, 540-700 nm, 780-975 nm, and 860-910 nm. High-phase-angle observations are of interest for studying the energy balance of Uranus, constraining the atmospheric scattering behavior, and understanding the planet as an analog for ice giant exoplanets. The new observations from New Horizons provide access to a wider wavelength range and different season compared to previous observations from both Voyager spacecraft. We performed aperture photometry on the New Horizons observations of Uranus to obtain its brightness in each photometric band. The photometry suggests that Uranus may be darker than predicted by a Lambertian phase curve in the Blue and Red filters. Comparison to simultaneous low-phase Hubble WFC3 and ground-based community-led observations indicates a lack of large-scale features at full-phase that would introduce variation in the rotational light curve. The New Horizons reflectance in the Blue (492 nm) and Red (624 nm) filters does not exhibit statistically significant variation and is consistent with the expected error bars. These results place new constraints on the atmospheric model of Uranus and its reflectivity. The observations are analogous to those from future exoplanet direct-imaging missions, which will capture unresolved images of exoplanets at partial phases. These results will serve as a "ground-truth" with which to interpret exo-ice giant data., Comment: Accepted for publication in the Planetary Science Journal, 11 pages, 6 figures

We present the mission concept "Mission to Analyze the UltraViolet universE" (MAUVE), a wide-field spectrometer and imager conceived during the inaugural NASA Astrophysics Mission Design School. MAUVE responds to the 2023 Announcement of Opportunity for Probe-class missions, with a budget cap of \$1 billion, and would hypothetically launch in 2031. However, the formulation of MAUVE was an educational exercise and the mission is not being developed further. The Principal Investigator-led science of MAUVE aligns with the priorities outlined in the 2020 Astrophysics Decadal Survey, enabling new characterizations of exoplanet atmospheres, the early-time light curves of some of the universe's most explosive transients, and the poorly-understood extragalactic background light. Because the Principal Investigator science occupies 30% of the observing time available during the mission's 5 yr lifespan, we provide an observing plan that would allow for 70% of the observing time to be used for General Observer programs, with community-solicited proposals. The onboard detector (THISTLE) claims significant heritage from the Space Telescope Imaging Spectrograph on Hubble, but extends its wavelength range down to the extreme UV. We note that MAUVE would be the first satellite in decades with the ability to access this regime of the electromagnetic spectrum. MAUVE has a field of view of 900" x 900" a photometric sensitivity extending to $m_{UV}\leq 24$, and a resolving power of $R\sim1000$. This paper provides full science and mission traceability matrices for this concept, and also outlines cost and scheduling timelines aimed at enabling a within-budget mission and an on-time launch., Comment: 21 pages, 15 figures. Published by the Publications of the Astronomical Society of the Pacific

This paper presents near-infrared spectropolarimetric and velocimetric observations of the young planet-hosting T Tauri star PDS 70, collected with SPIRou at the 3.6m Canada-France-Hawaii Telescope from 2020 to 2024. Clear Zeeman signatures from magnetic fields at the surface of PDS 70 are detected in our data set of 40 circularly polarized spectra. Longitudinal fields inferred from Zeeman signatures, ranging from -116 to 176 G, are modulated on a timescale of 3.008$\pm$0.006 d, confirming that this is the rotation period of PDS 70. Applying Zeeman-Doppler imaging to subsets of unpolarized and circularly polarised line profiles, we show that PDS 70 hosts low-contrast brightness spots and a large-scale magnetic field in its photosphere, featuring in particular a dipole component of strength 200-420 G that evolves on a timescale of months. From the broadening of spectral lines, we also infer that PDS 70 hosts a small-scale field of 2.51$\pm$0.12 kG. Radial velocities derived from unpolarized line profiles are rotationally modulated as well, and exhibit additional longer-term chromatic variability, most likely attributable to magnetic activity rather than to a close-in giant planet (with a 3sigma upper limit on its minimum mass of ~4 Mjup at a distance of ~0.2 au). We finally confirm that accretion occurs at the surface of PDS 70, generating modulated red-shifted absorption in the 1083.3-nm He i triplet, and show that the large-scale magnetic field, often strong enough to disrupt the inner accretion disc up to the corotation radius, weakens as the star gets fainter and redder (as in 2022), suggesting that dust from the disc more easily penetrates the stellar magnetosphere in such phases., Comment: MNRAS in press (20 pages, 12 figures, 5 tables)

Tungsten peaking is predicted in the core of a WEST plasma with total-f gyrokinetic simulations, including both collisional and turbulent transport. This prediction is validated with a synthetic diagnostic of the bolometry. Although nitrogen impurities are shown to reduce the neoclassical peaking of tungsten on-axis, the overall tungsten peaking increases when nitrogen impurities are present, as they reduce the turbulence screening off-axis. This finding is important for the plasma current ramp-up phase of ITER, where light impurities seeding will be desirable to achieve low temperatures at the plasma-facing components and reduce tungsten sputtering. It provides further argument for applying early ECRH heating to maintain margins on the core power balance. The neoclassical peaking factor is cross-verified between XGC and FACIT. The heat flux at separatrix and the heat load width are modeled by XGC and compared to WEST data., Comment: 15 pages, 17 figures

Theoretical studies have suggested that nonlinear effects can lead to "radio frequency condensation", which coalesces RF power deposition and driven current near the center of a magnetic island. It is predicted that an initially broad current profile can coalesce in islands when they reach sufficient width, providing automatic stabilization. Experimental validation of the theory has thus far been lacking. This paper proposes experiments on DIII-D for testing and refining the theory of the nonlinear effects.

A new axisymmetric equilibrium solver has been written, called FEQIS (Flexible EQuIlibrium Solver), which purpose is to be used inside integrated modeling of tokamak plasmas. The FEQIS code solves the Grad-Shafranov equation and the "circuit" equations for the external coils and passive conducting structures that are toroidally connected. The code has been specifically equipped with flexibility in choice of circuit connections, and a stripped-down numerical scheme for the solution of the Grad-Shafranov equation through a structure of multi-level simplifications which can be tested against the required accuracy.

This paper provides an empirical overview of the impact and practices of paying Article Processing Charges (APCs) by four nationally categorized groups of researchers in Argentina, Brazil, Mexico, and South Africa. The data was collected from 13,577 researchers through an online questionnaire. The analysis compares the practice of publishing in journals that charge APCs across different dimensions, including country, discipline, gender, and age of the researchers. The paper also focuses on the maximum amount APC paid and the methods and strategies researchers use to cover APC payments, such as waivers, research project funds, payment by coauthors, and the option to publish in closed access, where possible. Different tendencies were identified among the different disciplines and the national systems examined. Findings show that Argentine researchers apply for waivers most frequently and often use personal funds or international coauthors for APCs, with younger researchers less involved in APC payments. In contrast, Brazil, South Africa, and Mexico have more older researchers, yet younger researchers still publish more in APC journals. South African researchers lead in APC publications, likely due to better funding access and read and publish agreements. This study lays the groundwork for further analysis of gender asymmetries, funding access, and views on the commercial Open Access model of scientific dissemination., Comment: 33 pages, 6 figures, 12 tables

We used data from the Juno spacecraft to investigate both the spatial and temporal properties of Loki Patera on Io, acquired in two infrared bands between December 2022 and April 2024, at spatial resolutions ranging from 400 m to 15 km. Loki shows a thermal structure unlike other active lava lakes previously reported, with some brightening near the perimeter of the lake but lacking the continuous hot ring seen at other paterae. Modeling the slow rate of cooling suggests there is a significant volume of magma beneath the crust to provide the latent heat necessary to decelerate the cooling rate. A thermal propagation that may represent the signature of a resurfacing wave, going from the southwest of the lake to the north, was observed with a velocity of about 2-3 km per day. Data collected in 2024 may indicate the onset of a new resurfacing wave originating from a point source, rather than the foundering of a linear section of the crust. We also observed many small (about 3 km wide), closely spaced (about 10 km apart) islands that have persisted in the same locations for at least 45 years, since first being imaged by Voyager 1. The persistence of these islands challenges resurfacing models of Loki, as they have remained fixed - likely anchored to the lava lake floor - and have not noticeably changed in size, arguing against large-scale thermal erosion. The central island of Loki shows a few thermal structures associated with the fractures that cross the island, indicating that the fractures most likely contain molten lava.

For safety reasons, large language models (LLMs) are trained to refuse harmful user instructions, such as assisting dangerous activities. We study an open question in this work: does the desired safety refusal, typically enforced in chat contexts, generalize to non-chat and agentic use cases? Unlike chatbots, LLM agents equipped with general-purpose tools, such as web browsers and mobile devices, can directly influence the real world, making it even more crucial to refuse harmful instructions. In this work, we primarily focus on red-teaming browser agents, LLMs that manipulate information via web browsers. To this end, we introduce Browser Agent Red teaming Toolkit (BrowserART), a comprehensive test suite designed specifically for red-teaming browser agents. BrowserART is consist of 100 diverse browser-related harmful behaviors (including original behaviors and ones sourced from HarmBench [Mazeika et al., 2024] and AirBench 2024 [Zeng et al., 2024b]) across both synthetic and real websites. Our empirical study on state-of-the-art browser agents reveals that, while the backbone LLM refuses harmful instructions as a chatbot, the corresponding agent does not. Moreover, attack methods designed to jailbreak refusal-trained LLMs in the chat settings transfer effectively to browser agents. With human rewrites, GPT-4o and o1-preview-based browser agents attempted 98 and 63 harmful behaviors (out of 100), respectively. We publicly release BrowserART and call on LLM developers, policymakers, and agent developers to collaborate on improving agent safety

Methodologies for combining the accuracy of data-driven models with extrapolability of physics-based models are described and tested, for the task of building transport models of tokamak fusion reactors that extrapolate well to new operational regimes. Information from multiple physics simulations (the ASTRA transport code with gyro-Bohm and TGLF estimates for turbulent diffusion) as well as multiple distinct experiments (DIII-D and AUG tokamaks) are considered. Applications of the methodology to the task of commissioning and controlling a new reactor such as ITER are discussed.

Proper quality control (QC) is time consuming when working with large-scale medical imaging datasets, yet necessary, as poor-quality data can lead to erroneous conclusions or poorly trained machine learning models. Most efforts to reduce data QC time rely on outlier detection, which cannot capture every instance of algorithm failure. Thus, there is a need to visually inspect every output of data processing pipelines in a scalable manner. We design a QC pipeline that allows for low time cost and effort across a team setting for a large database of diffusion weighted and structural magnetic resonance images. Our proposed method satisfies the following design criteria: 1.) a consistent way to perform and manage quality control across a team of researchers, 2.) quick visualization of preprocessed data that minimizes the effort and time spent on the QC process without compromising the condition or caliber of the QC, and 3.) a way to aggregate QC results across pipelines and datasets that can be easily shared. In addition to meeting these design criteria, we also provide information on what a successful output should be and common occurrences of algorithm failures for various processing pipelines. Our method reduces the time spent on QC by a factor of over 20 when compared to naively opening outputs in an image viewer and demonstrate how it can facilitate aggregation and sharing of QC results within a team. While researchers must spend time on robust visual QC of data, there are mechanisms by which the process can be streamlined and efficient., Comment: 22 pages, 12 figures, 1 table, 6 supplemental figures

We propose a scaling law hypothesis for multimodal models processing text, audio, images, and video within a shared token and embedding space. Our framework predicts model performance based on modality-specific compression and tokenization efficiency, extending established scaling laws from text-based decoder models to mixed-modality systems. We explore whether leveraging more training data in multiple modalities can reduce the size of the multimodal model, enabling efficient deployment on resource-constrained devices.

Nuclear fusion could offer clean, abundant energy. However, managing the immense power exhausted from the core fusion plasma towards the divertor remains a major challenge. This is compounded in emerging compact reactor designs which promise more cost-effective pathways towards commercial fusion energy. Alternative divertor configurations (ADCs) are a potential solution to this challenge. In this work, we demonstrate exhaust control in ADCs for the first time, on MAST-U. We employ a novel diagnostic strategy for the neutral gas buffer which shields the target. Our work shows that ADCs tackle key risks and uncertainties in realising fusion energy: 1) an enlarged operating window which 2) improves exhaust control through the absorption of transients which can remove the neutral shield and damage the divertor, 3) isolation of each divertor from other reactor regions, enabling combined control. This showcases real-world benefits of alternative divertors for effective heat load management and control in reactors.

Mechanisms of plasmoid drift following massive material injection are studied via 3D non-linear MHD modelling with the JOREK code, using a transient neutral source deposited at the low field side midplane of a JET H-mode plasma to clarify basic processes and compare with existing theories. The simulations confirm the important role of the propagation of shear Alfv\'en wave (SAW) packets from both ends of the plasmoid (``SAW braking'') and the development of external resistive currents along magnetic field lines (``P\'egouri\'e braking'') in limiting charge separation and thus the $\mathbf{E}\times \mathbf{B}$ plasmoid drift, where $\mathbf{E}$ and $\mathbf{B}$ are the electric and magnetic fields, respectively. The drift velocity is found to be limited by the SAW braking on the few microseconds timescale for cases with relatively small source amplitude while the P\'egouri\'e braking acting on a longer timescale is shown to set in earlier with larger toroidal extent of the source, both in good agreement with existing theories. The simulations also identify the key role of the size of the $\mathbf{E}\times \mathbf{B}$ flow region on plasmoid drift and show that the saturated velocity caused by dominant SAW braking agrees well with theory when considering an effective pressure within the $\mathbf{E}\times \mathbf{B}$ flow region. The existence of SAWs in the simulations is demonstrated and the 3D picture of plasmoid drift is discussed.

The "density limit" is one of the fundamental bounds on tokamak operating space, and is commonly estimated via the empirical Greenwald scaling. This limit has garnered renewed interest in recent years as it has become clear that ITER and many tokamak pilot plant concepts must operate near or above the widely-used Greenwald limit to achieve their objectives. Evidence has also grown that the Greenwald scaling - in its remarkable simplicity - may not capture the full complexity of the disruptive density limit. In this study, we assemble a multi-machine database to quantify the effectiveness of the Greenwald limit as a predictor of the L-mode density limit and identify alternative stability metrics. We find that a two-parameter dimensionless boundary in the plasma edge, $\nu_{*\rm, edge}^{\rm limit} = 3.0 \beta_{T,{\rm edge}}^{-0.4}$, achieves significantly higher accuracy (true negative rate of 97.7% at a true positive rate of 95%) than the Greenwald limit (true negative rate 86.1% at a true positive rate of 95%) across a multi-machine dataset including metal- and carbon-wall tokamaks (AUG, C-Mod, DIII-D, and TCV). The collisionality boundary presented here can be applied for density limit avoidance in current devices and in ITER, where it can be measured and responded to in real time., Comment: 27 pages, 9 figures